View original document

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

J U LY /A U G U S T 1996

Cletus C. Coughlin is associate director of research and Patricia S. Pollard is an economist at the Federal Reserve Bank of St. Louis. Jerram
Betts provided research assistance.

A Question of
Measurement:
Is the Dollar
Rising or
Falling?

Board, the U.S. dollar fell in value by 62
percent between March 1985 and December 1995.2 In contrast, the index produced
by the Federal Reserve Bank of Dallas
shows the dollar rising in value by 60 percent during the same period.
Even when the indexes show the dollar
moving in the same direction, they generally do not agree on the overall magnitude
of that change. Why don’t these indexes
provide a consistent view of changes in the
value of the dollar? This article answers
this question by examining the way in
which exchange rate indexes are constructed. We begin by exploring the basic
issues of constructing effective exchange
rates using the six indexes shown in Figure
1 for illustration. After discussing the differences in constructing these indexes, we
examine some factors that might account
for the contrasting views of the dollar by
focusing on two specific indexes—the Federal Reserve Board and the Federal Reserve
Bank of Dallas indexes.

Cletus C. Coughlin and
Patricia S. Pollard

I

n March 1985 one U.S. dollar could buy
258 Japanese yen and 0.21 Mexican
pesos. In December 1995 the same dollar could buy only 102 yen, but could now
buy 7.7 Mexican pesos. Though the
change in the value of the dollar against
each of these currencies was exceptionally
large, the behavior of the dollar—rising
against one currency and falling against
another—was not uncommon. Over the
past 10 years the dollar has appreciated
against many currencies and depreciated
against others. How then can one determine what has happened to the overall
value of the dollar? Is the dollar stronger
or weaker than it was 10 years ago? To
begin answering this question, economists
construct effective exchange rate indexes.
Effective exchange rates, commonly
termed trade-weighted exchange rates, measure the average foreign exchange value of
a country’s currency relative to a group of
other currencies.1 Unfortunately, looking at
effective exchange rate indexes may not
provide a consistent answer to the preceding questions. The effective exchange value
of the dollar as measured by six commonly
used indexes is shown in Figure 1. According to four of these indexes, the dollar has
fallen in value since March 1985, whereas
two other indexes show a rise in the value
of the dollar since March 1985. For example, according to the effective exchange
rate index produced by the Federal Reserve

CONSTRUCTING EFFECTIVE
EXCHANGE RATE INDEXES
The construction of effective exchange
rate indexes requires a number of decisions.3 Because many of the decisions have
more than one defensible alternative, it is
not surprising that a number of effective
exchange rate indexes are used. Six decisions are examined: (1) which formula is
used to calculate the average, (2) which
foreign currencies are used in the calculation, (3) which measure of economic activity is used as the basis for weighing the importance of individual currencies, (4) how
to calculate the weights for individual currencies, (5) the base period for calculating
the weights, and (6) the base period for
calculating exchange rate changes. These
decisions are illustrated with specific references to how six well-known effective exchange rate indexes are constructed. These
indexes are identified by their producers—

FEDERAL RESERVE BANK

3

OF

S T. L O U I S

1

Effective exchange rate indexes
were developed by the International Monetary Fund. The
seminal work was by Hirsch
and Higgins (1970).

2

For all indexes discussed in this
article, percentage changes are
calculated on a logarithmic
basis. Thus the percentage
change in an index that increases from 100.0 to 111.2
is the natural logarithm of the
ratio of 111.2 to 100 or 10.6
percent.

3

The issues involved in constructing effective exchange
rate indexes have been discussed by many authors, including Rhomberg (1976),
Rosensweig (1987), and
Turner and Van ‘t dack.
(1993).

J U LY /A U G U S T 1996

dollar could buy 25 units of Currency x. In
the second year one dollar could buy 50
units of Currency x, and in the third year a
dollar could buy 100 units of Currency x.
With respect to Currency y, one dollar
could buy 40 units in the first year, 20
units in the second year, and 10 units in
the third year. The dollar rose in value
against Currency x—over time one dollar
could buy more and more units of this currency. In contrast, the dollar fell in value
against Currency y—over time one dollar
could buy fewer and fewer units of this
currency. Note that compared with the first
year, one dollar could buy twice as many
units of Currency x and half as many units
of Currency y in the second year, and four
times as many units of Currency x and onequarter as many units of Currency y in the
third year.
What happened to the overall value
of the dollar? There are two methods of
calculating an average value for the dollar: an arithmetic mean or a geometric
mean. Each method compares the effective value of the dollar with its value in a
given period, for example, relative to the
first year. An arithmetic mean computes a
simple average. In Year 1 the effective exchange rate using the arithmetic mean is

Figure 1

Effective Exchange Rates

Index

(March 1985=1)
200
180
160
140
120
100
80
60
40
20

Board Dallas Atlanta

76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
Year

Selected Effective Exchange Rates
March 1985=100
140
Morgan B Morgan N IMF
120
Index

100
80
60
40

4

76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
Year

Following this general principle
will not necessarily mean that
the constructed exchange rate
measure will generate superior
results when used in a specific
case. See Belongia (1986) for
an empirical demonstration supporting such a conclusion in the
context of U.S. agricultural exports. See Deephouse (1985)
and Hooper and Morton
(1978) for an elaboration of
the uses of effective exchange
rate indexes.

e
e 
1  ex , 1x, +e y,y1,  =
e + e  =
2  ex , 1x, e y,y1, 

Federal Reserve Board, J.P. Morgan (broad
and narrow), International Monetary Fund
(IMF), Federal Reserve Bank of Dallas, and
Federal Reserve Bank of Atlanta. The
movement of these indexes over time is
presented in Figure 1, and a summary of
their construction characteristics is provided in Table 1. In sorting through the
various choices in constructing an index, it
may be helpful to keep in mind a general
principle: The use of the index should
guide its construction.4


1  25+

+
2  25


40 = ,
 =1 ,

40

where ex,1 is the Currency x/dollar exchange rate in Year 1, and ey,1 is the Currency y/dollar exchange rate in Year 1. In
Year 2 the effective exchange rate using
the arithmetic mean is


20
1  eexx, ,2 + eeyy, , 2 = 1  50

 =  + + =  = 1, .25 ,
+
2 eexx,,1 eeyy, ,1  2  25 40 

where ex,2 and ey,2 are the Currency x/dollar
exchange rate and the Currency y/dollar exchange rate, respectively, in Year 2. Similarly,
in Year 3 the effective exchange rate using
the arithmetic mean is

Which Formula?
Suppose the world has three currencies—the dollar, Currency x and Currency y.
Further suppose that in the first year one

FEDERAL RESERVE BANK

4

OF

S T. L O U I S

J U LY /A U G U S T 1996

Table 1

Construction Features of Effective Exchange Rates for the Dollar
Producer

Number of
Countries

Trade-Weight
Period

1967–present
1970–1986
1987–present
1970–1986
1987–present
1957–present

10
15
18
44
44
20

1972–1976
1980
1990
1980
1990
1989–1991

1976–present

128

1973–present

18

Three-year moving
average
1984

Years Covered

Federal Reserve Board
J.P. Morgan (narrow)
J.P. Morgan (broad)
International
Monetary Fund
Federal Reserve
Bank of Dallas
Federal Reserve
Bank of Atlanta


1  e x , 3 e x,e y, 3ey,  1  100
10 
+ = =  ++  = 2.125, ,
+

2  e x , 1 e x, e y, 1ey,  2  25 40 

where ex,3 and ey,3 are the Currency x/dollar
exchange rate and the Currency y/dollar exchange rate, respectively, in Year 3. The resulting number in each year is generally
multiplied by 100 to create an easily usable
index. Thus the effective exchange rate index for the three years is 100, 125, and
212.5.
The geometric mean in Year 1, again
using the first year as the base year, is
1

1

 e x ,1 e x, e y, 1ey,2   25

40  2
 +×  = 1. .
 e  e+ e× e = = 25
 x ,1 x, y, 1 y,   40 
In the Year 2 the geometric mean is
1

1

 e x , 2  e ex,y, 2 e 2y,   50 20  2
 e +e e × ey, =  =25 + ×40  == 1 ..

 x , 1  x, y , 1   
In Year 3 the geometric mean is
1

1

 e x , 3 e xe, y, 3 e 2y,   100

10  2
 e +e e × e =  =25 + ×40  == 1. ..
 x , 1 x, y, 1  y, 

Weighting Scheme
Multilateral
Double (manufactures)
Double (manufactures)
Double (manufactures)
Double (manufactures)
Double (manufactures)
Bilateral
Bilateral

Multiplying the resulting number in each
year by 100 produces the following index for
the three years: 100, 100, 100.
Using the arithmetic mean, the effective
value of the dollar rose over the three-year
period, whereas using the geometric mean,
the effective value of the dollar was unchanged. The result based on the geometric
mean seems more reasonable, given that the
rise in the value of the dollar against Currency x is offset by the fall in the value of
the dollar against Currency y. The arithmetic
mean created an upward bias.5 The Board of
Governors of the Federal Reserve System,
when it switched from using an arithmetic
mean to a geometric mean to construct its
effective exchange rate index for the dollar,
noted that “as currencies diverged from each
other over time, changes in currencies that
rose against the dollar had a reduced impact
on the index while changes in currencies
that fell against the dollar had an increased
impact on the index. As a result, arithmetic
averaging imparted a systematic upward bias
to the measurement of changes in the dollar’s average exchange value.”6
Because of the bias inherent in an index
based on arithmetic averaging, all the effective exchange rate indexes shown in Figure
1 use a geometric averaging technique. Of
the six decisions involved in constructing an
effective exchange rate index, this choice of

FEDERAL RESERVE BANK

5

OF

S T. L O U I S

5

It is not mandatory that the direction of the bias be upward.
If Year 3 had been used as the
base year, the index using the
arithmetic average would be
212.5, 125, 100 and the
index using geometric averaging would be 100, 100, 100.
In this example, arithmetic averaging would have created a
downward bias.

6

See Board of Governors
(1978), p. 700.

J U LY /A U G U S T 1996

a geometric average is the only one on
which there is consensus.
The generic formula, using geometric
averaging, for the value of the effective exchange rate index at time t is
(1)

7

See Hooper and Morton
(1978).

8

Whether indexes with a broad
range of currencies are superior
to those using a small range of
currencies is an empirical question. See Batten and Belongia
(1987) for an empirical study
of U.S. trade flows indicating
that measures based on more
currencies performed no better
than the measures based on
fewer currencies.

9

J.P. Morgan and the IMF produce effective exchange rate
indexes for each of the currencies included in the U.S. dollar
indexes.

10

For more on the choice of currencies in the Atlanta index, see
Rosensweig (1986a and b).

11

The 26 countries included in
J.P. Morgan’s broad, but not its
narrow, index are Ireland, New
Zealand, Turkey, Argentina,
Brazil, Chile, Colombia,
Ecuador, Mexico, Peru,
Venezuela, Hong Kong, Indonesia, South Korea, Malaysia,
Philippines, Singapore, Taiwan,
Thailand, India, Kuwait, Morocco, Nigeria, Pakistan, Saudi
Arabia, and South Africa.

12

Cox (1986) stressed that the
index contained all U.S. trading
partners; however, the index
contains few currencies from
Eastern European countries and
countries that were formerly
part of the Soviet Union.

n
eit
Indext = 100Π }
eib
i=1

1 2

plus Switzerland. These countries were selected for several reasons.7 First, each
country has a well-developed foreign exchange market with exchange rates that depend primarily on the supply and demand
decisions of private individuals and firms.
Second, these countries are involved in the
majority of U.S. trade and capital flows.
Third, many of the countries excluded
from the index either attempt to keep their
currencies pegged to an included currency
or use one of the included currencies for
their international transactions.
The countries whose currencies are included in the index produced by the Federal Reserve Board are located in Europe,
except for Canada and Japan. Clearly, this
index includes the major traded currencies
and consequently allows an assessment of
changes in the value of the U.S. dollar relative to the other major currencies. The
other five indexes discussed here use the
10 currencies in the Board’s index, but they
add other currencies as well.8 For example,
the narrow index produced by J.P. Morgan
adds currencies from seven European
countries—Austria, Denmark, Finland,
Greece, Norway, Portugal and Spain—plus
Australia. The currencies of Finland,
Greece, and Portugal did not appear in the
index until 1987. The IMF index adds the
currencies of Ireland and New Zealand to
the J.P. Morgan narrow index. The IMF
index therefore contains the currencies of
all the major industrialized countries.9 The
Atlanta index adds the currencies of Taiwan, Hong Kong, South Korea, Singapore,
and China, as well as those of Australia,
Spain, and Saudi Arabia, to the Board’s
index. The addition of the currencies of the
first five countries is justified by the shifting pattern of U.S. trade toward developing
countries in Asia.10 In addition to a narrow
index for the United States, J.P. Morgan
produces a broad index that uses the currencies of most member countries of the
Organization for Economic Cooperation
and Development plus numerous developing countries.11 The ultimate in inclusiveness is the index produced by the Federal
Reserve Bank of Dallas, which currently
includes 128 currencies.12

wit

,

where Π is the product over the n foreign currencies in the index, eit is the
number of units of Currency i per dollar
at time t; eib is the number of units of
Currency i per dollar in the base period;
and wit is the weight assigned to Currency i at time t.
In the above example, each currency
was given equal weight in each period,
wit = 1/2 and the base period was Year 1.
In actually constructing an exchange rate
index, developers must make numerous
decisions involving the currencies included, the weights for the currencies, and
the base periods. An elaboration of the key
decisions is provided below.

Which Currencies?
Ideally, an effective exchange rate for
the dollar should include all currencies
for which the dollar is exchanged. Such
an ideal, however, is tempered by the reality that the construction of the index requires timely, reliable data. As a result,
most indexes are limited to the currencies
of the principal industrial economies.
Table 1 shows that most indexes use data
on the dollar relative to the currencies of
between 10 and 20 countries. The major
exceptions are the broad index produced
by J.P. Morgan that uses the currencies of
44 countries relative to the dollar and the
index produced by the Federal Reserve
Bank of Dallas that currently uses the
currencies of 128 countries.
The index produced by the Federal Reserve Board uses data on the dollar relative
to the currencies of the other nine members of the Group of Ten—Belgium,
Canada, France, Germany, Italy, Japan,
Netherlands, Sweden, United Kingdom—

FEDERAL RESERVE BANK

6

OF

S T. L O U I S

J U LY /A U G U S T 1996

termine the weights. Both the absolute levels and the rapid growth rates of international capital flows suggest that capital
flows might currently be a more important
determinant of exchange rates than trade
flows. Thus using capital flows, the currencies of countries with larger investment
and portfolio flows are more important in
the determination of the value of the dollar
than are the currencies of countries with
smaller investment and portfolio activity.
Even though such a calculation is reasonable on theoretical grounds, no major producer of effective exchange rates uses capital flows to construct its measures.13
A key reason trade is used for weighting purposes is that, although trade data are
subject to errors, they are much easier to
obtain on a timely basis than capital flows.
Different indexes, however, use different
measures of international trade. Generally
speaking, most indexes are constructed
using total merchandise trade and do not
include services, which have tended to increase rapidly in recent years. The indexes
produced by J.P. Morgan and the IMF, however, use only trade in manufactures.

Which Measure of
Economic Activity?
Deciding how many countries to include in the index also requires decisions
concerning how much importance should
be attached to the currency from a particular country. In other words, the relative
importance of a currency is determined by
its weight in the average. Before determining the weight of a particular currency, researchers must decide which measure of
economic activity is used in the calculation of the weights.
Because effective exchange rate indexes are most often constructed to measure changes in a country’s international
competitiveness, generally some measure
of international trade is used to calculate
the weights. For this reason, effective exchange rates are frequently termed tradeweighted exchange rates. International
trade, however, is not the only measure of
international economic activity that could
be used. The exchange value of the dollar
is determined by supply and demand
forces involving the international exchange of goods, services, and assets. Individuals, firms, and governments demand
(buy) dollars in foreign exchange markets
to purchase goods, services, or assets denominated in U.S. dollars. Likewise, individuals, firms, and governments supply
(sell) dollars in foreign exchange markets
to purchase goods, services, or assets denominated in foreign currencies. For example, a U.S. auto dealer wanting to import BMWs must first obtain German
marks and so supplies dollars and demands marks. Any country wanting to import petroleum must pay in U.S. dollars
and so must first exchange its own currency for dollars, supplying its currency
and demanding dollars. A Japanese investor who wants to buy U.S. Treasury securities must first obtain U.S. dollars and
so supplies yen and demands dollars.
Though trade flows are used to calculate the weights given to each currency in
an effective exchange rate index, based on
international financial movements, one
could use international capital flows to de-

Which Weighting Method?
Another issue in weighting the importance of a specific currency involves the
selection of a weighting scheme. If the effective exchange rate index is to reflect
changes in a country’s international competitiveness, then ideally the weights
should be chosen to reflect the responsiveness of a country’s trade flows to changes
in exchange rates. A theoretically based
index was previously produced by the
IMF: the Multilateral Exchange Rate
Model (MERM) index. In the U.S. dollar
MERM index, for example, the weight
given to each currency was chosen so that
any combination of changes in the currencies against the dollar leading to a one percent change in the index would have the
same effect on the U.S. trade balance (over
a 2-3 year period) as a one percent change
in the dollar against each currency in the
index. Estimation of the weights required
the use of an econometric model incorpo-

FEDERAL RESERVE BANK

7

OF

S T. L O U I S

13

See Ott (1987) for a more extensive discussion and illustration of a capital-weighted exchange rate.

J U LY /A U G U S T 1996

rating information on price elasticities, exchange rate effects on domestic prices, and
the policy response of the economy. Concerns about the unreliability of the model
determining the weights led to the abandonment of the MERM and similarly constructed indexes.14
Three other methods of weighting remain in use: bilateral, multilateral, and double weights.15 With bilateral weighting, each
country is weighted by the proportion of its
share of the total trade flows to and from
the United States of the countries used to
construct the index. Thus the weight for
Country i is simply the sum of U.S. exports
to and imports from Country i divided by
the sum of U.S. exports to and imports from
all countries included in the index. Assuming that n countries are used to construct
the index, the weight for Country i is:

(2)

tween two countries in countries outside of
their domestic markets. For example, a
change in the Japanese yen-U.S. dollar exchange rate can affect relative prices of
Japanese goods, American goods, and
goods from other countries besides Japan
and the United States, such as Canada.
The multilateral approach used in the construction of the index produced by the Federal Reserve Board seems more suitable for
accounting for these third-country effects.
On the other hand, it is possible that the
multilateral weighting approach gives too
much weight to nations that trade more extensively with each other than with the
United States. For example, European
Community countries that trade extensively with each other are likely to receive
higher-than-warranted weights in the
construction of an index for the United
States. A possible result in the case of an
effective exchange rate for the United States
would be that Canada, the largest U.S. trading partner, would be weighted less than
warranted. In this case, a bilateral weighting approach that is used in the indexes
produced by the Federal Reserve Bank of
Dallas and the Federal Reserve Bank of Atlanta might be more appropriate.
The double weighting method, which
is used in the indexes produced by the
IMF and J.P. Morgan, attempts to combine
the advantages of both the bilateral and
multilateral weighting schemes: recognition of competition in third markets and
the strength of links between particular
trading partners. In addition, the double
weighting method recognizes the competitive position of domestic producers
of import substitutes and therefore requires information on production for
local consumption as well as on trade
flows.17 In the dollar index, the weights
reflect both the competition U.S. ex-porters
face from other countries’ exporters and
from the local countries’ producers.

USXi + USMi
n
wi = }}
,
^(USXi + USMi)
i=1

14

Turner and Van ‘t dack (1993)
provide a good overview of the
construction and problems associated with the MERM index.

15

Bilateral weights were used in
the original work on effective
exchange rates, see Hirsch and
Higgins (1970).

16

To simplify the discussion we
have omitted all references to
time. Obviously, the trade
flows cover a particular period
and the weight for a country
pertains to a particular period.
As indicated by equation 1 and
discussed in the next section,
the weight for a country may
change over time.

17

See Hargreaves (1993) for details on how the J. P. Morgan
index is constructed. Turner and
Van ‘t dack (1993) provide a
general analysis of the double
weighting method.

where USXi is the exports from the United
States to Country i and USMi is the imports
of the United States from Country i.16
With multilateral weighting, each
country is weighted by the proportion of
its share of total trade flows throughout
the world. Thus the weight for each Country i is the sum of Country i’s worldwide
exports and imports divided by the sum of
the worldwide exports and imports of all
the countries included in the index. Once
again, assuming that n countries are used
to construct the index, the weight for
Country i is:

(3)

WXi + WMi
wi = }}
,
n
(WX
+
WM
)
i
^ i
i=1

where WXi is the worldwide exports of
Country i and WMi is the worldwide imports of Country i.
Neither alternative is obviously superior. The multilateral weighting approach
attempts to capture the competition be-

Which Base Period for Weights?
The fifth major issue in the construction of an effective exchange rate is the
choice of a base period for the trade flows

FEDERAL RESERVE BANK

8

OF

S T. L O U I S

J U LY /A U G U S T 1996

on which the weights are based. The index
may use fixed weights, weights that are
updated periodically, or weights that are
updated annually. For example, the Federal
Reserve Board’s index uses fixed weights
that have remained unchanged; the J.P.
Morgan indexes use different weights for
the period from 1970 to 1986 and the period from 1987 to the present; and the index produced by the Federal Reserve Bank
of Dallas uses a three-year moving average
to continually update its weights.18 If fixed
weights are used, then researchers must decide which year or years should be used.
For example, the Federal Reserve Bank of
Atlanta index uses 1984 trade figures, the
Federal Reserve Board index uses trade
data from 1972 to 1976, and the IMF index
uses trade data from 1989 to 1991.
The existence of various base periods
suggests that there is no obviously superior
base period. Fixing the base period for the
trade weights means that the index does
not incorporate the effect of changing
trade patterns. Thus a shifting pattern of
trade raises the possibility that a fixedweight index becomes a less reliable exchange rate measure over time. On the
other hand, a potential problem stemming
from updating the weights annually is that
the effects of exchange rate changes may
be confounded with changes caused by
shifting weights in the index. It is possible,
because of shifts in trade shares, that an effective exchange rate may change
even if no individual exchange rate
changes.
Table 2 illustrates this point. The
upper half of the table shows the results
of calculating a hypothetical tradeweighted exchange rate index for the U.S.
dollar assuming fixed weights for each
currency based on trade shares at some
point. The weight for Country 1 is 0.7,
whereas the weight for Country 2 is 0.3.
The lower half of the table shows the results of calculating a hypothetical tradeweighted exchange rate index for the U.S.
dollar assuming that the weights given to
each currency are updated annually. In
the example, the weight for Country 1
declines from 0.7 in Year 1 to 0.3 in Year 7,

whereas the weight for Country 2 increases
from 0.3 in Year 1 to 0.7 in Year 7.
Between Year 5 and Year 6, the value
of the dollar was unchanged against both
currencies as 61 units of Country 1’s currency and 17 units of Country 2’s currency
could be traded for one U.S. dollar in each
year. The index calculated using fixed
weights shows no change in the effective
exchange value of the dollar. For example,
assuming that the effective exchange rate
in Year 1 equals 100, then the rate in both
Year 5 and Year 6 is 144.4. When weights
are updated often, however, the effective
exchange value of the dollar does change.
For example, assuming that the effective
exchange rate in Year 1 equals 100, then
the rate in Year 5 is 93.3 and the rate in
Year 6 is 78.4.
Thus changes in an index with
weights that are updated annually always
leave doubt as to whether changes in the
index reflect exchange rate changes or
shifting trade weights. On the other
hand, if trade patterns shift, then the use
of fixed weights may cause the index to
produce misleading signals. This is
highly likely over long periods. A compromise is to change the weights periodically; however, it is not obvious how frequently weights should be changed.

Which Base Period
for Exchange Rates?
The effective exchange rate index
shown in Equation 1 calculates changes in
the exchange rate of the domestic currency
(for our purposes the U.S. dollar) relative to
each foreign currency from a base exchange
rate. The Federal Reserve Board uses the
March 1973 exchange rates as the base
rates.19 The Federal Reserve Bank of Atlanta
uses 1980. The Federal Reserve Bank of
Dallas uses the exchange rate averages for
first quarter 1985 as the base. The IMF and
J.P. Morgan use the exchange rate averages
for 1990 as the base. As Equation 1 indicates,
the index in the base period equals 100.
The creation of effective exchange rate
indexes differs from that of most price in-

FEDERAL RESERVE BANK

9

OF

S T. L O U I S

18

For example, trade data for
1992–94 is used for calculating the index in 1995.

19

This period reflects the start of
the flexible exchange rate era.

J U LY /A U G U S T 1996

Table 2

Exchange Rate Indexes: Alternative Updating Procedures for Weights*
Fixed Trade Weights
Exchange
Rates

Weights

Percent Change
in Index

Index

Year

e1

e2

w1

w2

Year 1 5 100

Year 7 5 100

Year 1 5 100

Year 7 5 100

1
2
3
4
5
6
7

25
32
39
49
61
61
70

40
32
26
21
17
17
13

0.7
0.7
0.7
0.7
0.7
0.7
0.7

0.3
0.3
0.3
0.3
0.3
0.3
0.3

100
111.2
120
132
144.4
144.4
146.7

68.1
75.8
81.7
90
98.4
98.4
100

—
10.6
7.6
9.6
9
0
1.6

—
10.6
7.6
9.6
9
0
1.6

Annually Updated Trade Weights
Exchange
Rates

Weights

Percent Change
in Index

Index

Year

e1

e2

w1

w2

Year 1 5 100

Year 7 5 100

Year 1 5 100

Year 7 5 100

1
2
3
4
5
6
7

25
32
39
49
61
61
70

40
32
26
21
17
17
13

0.7
0.65
0.6
0.5
0.45
0.35
0.3

0.3
0.35
0.4
0.5
0.55
0.65
0.7

100
108.6
109.9
101.4
93.3
78.4
62

68.1
82.4
92.9
106.3
108.9
113.5
100

—
8.2
1.2
28
28.4
217.5
223.4

—
19
12
13.5
2.4
4.1
212.6

* Note that e = foreign currency per dollar. Percentage changes are calculated on a logarithmic basis from the preceding year to the current year.

20

This issue is explored extensively in Coughlin, Pollard and
Betts (1996).

example in Table 2 can be used to illustrate this problem. Two versions of the
fixed trade weights and annually updated
trade weights indexes are calculated. One
version uses the exchange rates in Year 1
as the base rates. The other version uses
the exchange rates in Year 7 as the base
rates. When the trade weights are fixed,
changing the base year does not affect the
percentage change in the exchange rate
index. As shown in the last two columns
of the top panel of Table 2, the percentage
change in the effective exchange rate between any two years is the same regardless
of whether Year 1 or Year 7 is used as the
base year. As shown in the top panel of
Figure 2 under either base year for the exchange rate index, the index indicates an
appreciation of the dollar through Year 5, a

dexes in the use of two base periods. For
example, in the consumer price index the
base period for prices is exactly the same
as the base period for quantities. In effective exchange rate indexes the base periods
for weights and for exchange rates are generally different. The Atlanta index, for example, uses 1984 as the base period for the
trade data used to construct the weights
but uses first quarter 1985 as the base period for exchange rates.
The choice of the base exchange rate
period is irrelevant to the picture of the
dollar’s strength or weakness as measured
by indexes with fixed trade weights. When
the weights are updated annually, however,
the calculated percentage changes in the
value of the dollar become sensitive to the
base period for the exchange rates.20 The

FEDERAL RESERVE BANK

10

OF

S T. L O U I S

J U LY /A U G U S T 1996

constant value of the dollar from Year 5 to
Year 6, and a slight appreciation of the
dollar in Year 7.
The effective exchange value of the
dollar, however, is affected by the choice
of the base period for the exchange rate
when the trade weights are updated annually. As shown in the bottom halves of
Table 2 and Figure 2, if exchange rates in
Year 1 are used as a base, the effective exchange value of the dollar appreciates until
Year 3 and depreciates thereafter. If exchange rates in Year 7 are used as the base,
the effective exchange value of the dollar
rises through Year 6 and falls in Year 7.
Note that whereas the value of the dollar
is constant between Year 5 and Year 6
using fixed trade weights, when the trade
weights are continuously updated, the effective exchange rate index indicates
either a depreciation or an appreciation of
the dollar, depending on the base period
for the index.

Figure 2

Exchange Rate Indexes:
Fixed Weights
Using Different Base Years for the Exchange Rates
160
140
Index

120
100
80
60

Year 1=100 Year 7=100

1

2

3

4
Year

5

6

7

5

6

7

Exchange Rate Indexes
Annually Updated Weights
Using Different Base Years for the Exchange Rates
120
110
100
Index

WHAT ACCOUNTS FOR
DIFFERENCES IN THE
EXCHANGE RATE INDEXES?

90
80
70
Year 1=100 Year 7=100

60

Because exchange rates indexes are
constructed differently, it is not surprising
that the picture they give of the value of
the dollar may differ. The previous section
explained the choices creators of effective
exchange rate indexes face in designing an
index. This section concentrates on two
popular indexes––the Federal Reserve
Board (Board) index and the Federal Reserve Bank of Dallas (Dallas) index––to
illustrate which factors are the most important in accounting for differences in the behavior of the two indexes. As Figure 1
shows, these two indexes were qualitatively similar between January 1976 and
March 1985 but differed sharply between
March 1985 and December 1995. According to Table 3, during the early period the
Board index showed a 43 percent appreciation of the U.S. dollar, whereas the Dallas
index showed a substantially larger appreciation of the dollar, 77 percent. During the
later period the Board index showed a 62

50

1

2

3

4
Year

percent depreciation of the dollar. In sharp
contrast, the Dallas index showed a 60 percent appreciation of the dollar. Over the
sample period 1976–95 there was little
correlation between the two indexes, as
shown by the correlation coefficient of
20.27 in Table 4. In the early period the
indexes were highly positively correlated
(0.91), but exhibited a negative correlation
(20.50) in the later period.
The construction of the Board and
Dallas indexes differs in three aspects: the
method used to calculate the trade
weights, the base period for the trade
weights, and the choice of currencies in
each index.21 The Board index uses multilateral trade shares, whereas the Dallas
index uses bilateral trade shares. The

FEDERAL RESERVE BANK

11

OF

S T. L O U I S

21

The Board and Dallas indexes
also differ in their choice of
base period used for their exchange rates. To eliminate any
problems caused by this difference, we recalculated the
Board index using the March
1985 exchange rates as the
base rates.

J U LY /A U G U S T 1996

Table 3

Exchange Rate Changes in the Various Constructed
Trade-Weighted Exchange Rate Indexes (in percent)
Period

Board

Dallas

BilBoard

MupBoard*

BupBoard

CmBoard

CmupBoard

1976.01–1995.12
1976.01–1985.03
1985.03–1995.12

219
43
262

137
77
60

214
30
244

217
42
259

217
30
248

28
46
218

58
47
11

* The data period for the MupBoard index ends in December 1994.

Table 4

Correlations Among Trade-Weighted Exchange Rate Indexes
Correlation with the Board Index
Period

Dallas

BilBoard

MupBoard*

BupBoard

CmBoard

CmupBoard

1976.01–1995.12
1976.01–1985.03
1985.03–1995.12

20.27
0.91
20.5

0.98
0.99
0.99

1
1
1

0.97
0.99
0.99

0.52
0.97
0.94

0.11
0.97
0.02

Correlation with the Dallas Index
Period

Board

BilBoard

MupBoard*

BupBoard

CmBoard

CmupBoard

1976.01–1995.12
1976.01–1985.03
1985.03–1995.12

20.27
0.91
20.5

20.39
0.93
20.51

20.21
0.91
20.47

20.45
0.93
20.52

0.61
0.97
20.26

0.91
0.97
0.81

* The data period for the MupBoard index ends in December 1994.

weight assigned to each currency in the
Board index is fixed, whereas the weights
in the Dallas index are updated annually.
Specifically, the weights used in the Board
index were determined by the average
trade share of each country whose currency is included in the index for the period 1972–76. In contrast, in the Dallas
index, the weights used in a given year are
based on the average trade shares over the
prior three-year period. Last, the currencies of 10 countries are used in the Board
index, whereas the currencies of 128
countries are used in the Dallas index.
This section examines the importance
of each of these three aspects in accounting for the differences between the two indexes. It does so by creating five variations
on the Board index—BilBoard, MupBoard,
BupBoard, CmBoard, and CmupBoard—

shown in Figure 3. Each variation modifies the construction of the Board index so
that it is more closely in accord with the
Dallas index. These new indexes are used
to determine what causes the differences
between the Board and the Dallas indexes.
Table 5 presents an overview of these
five indexes, comparing them with the
Board and the Dallas indexes. The BilBoard
index is constructed using the same 10 currencies as in the Board index and the fixed
weights based on 1972–76 trade shares of
each country. However, whereas the Board
index uses the world trade of each country
to determine the weight given to its currency in the index, the BilBoard index uses
only the bilateral trade flows of the 10
countries with the United States. Contrasting this index with the Board and Dallas indexes allows us to determine the impor-

FEDERAL RESERVE BANK

12

OF

S T. L O U I S

J U LY /A U G U S T 1996

tance of the multilateral/bilateral trade
share choice in explaining the differences
between the latter two indexes.
The MupBoard index differs from
the Board index solely in the type of the
base period for the weights given to each
currency. Trade weights in the MupBoard
index are updated annually, using a threeyear moving average as in the Dallas
index. The MupBoard index can be contrasted with the Board and Dallas indexes
to determine the importance of the updating of weights in accounting for the differences between the latter two indexes.
The remaining difference between the
Board and Dallas indexes is the choice of
currencies used in each index. We created
three variations on the Board index to examine the importance of currency choice.
First we created BupBoard, an index that
was identical to the Dallas index except
that only the ten currencies used in the
Board index were included in its calculation. Thus any differences in the behavior
of the BupBoard and Dallas indexes could
be attributed to the difference in currency
choice between the Board and Dallas indexes. To further explore the importance
of currency choice, we added the currencies of China and Mexico to a bilateral–
trade share version of the Board index.
Mexico was chosen because it has consistently been the most important U.S. trading
partner excluded from the Board index.
China is currently the next most important
trading partner missing from the Board index. Its relative importance, as shown in
Table 6, has grown substantially over the
last 20 years. In 1976 the Chinese yuan received a weight of only 0.4 percent in the
Dallas index, but its weight rose to 3.9 percent by 1995. Using the Chinese yuan and
Mexican peso, we created two more indexes. In the CmBoard index, the weights
given to each of the 12 currencies are determined by each country’s share of trade
with the United States. This index therefore differs from the Board index in two
ways: it includes China and Mexico and
uses bilateral trade shares. The CmupBoard
index is constructed in the same manner as
the CmBoard index except that the weights
assigned to each currency are updated an-

Table 5

Overview of Variations on
the Board and Dallas Indexes*
Index

Trade
Shares

Base Period
for Weights

Currencies

Board
BilBoard
MupBoard
BupBoard
CmBoard
CmupBoard
Dallas

Multilateral
Bilateral
Multilateral
Bilateral
Bilateral
Bilateral
Bilateral

Fixed
Fixed
Updated annually
Updated annually
Fixed
Updated annually
Updated annually

10
10
10
10
12
12
128

* Note that the shaded cells highlight the differences from the Board index.

Table 6

Weights for the 10 Highest
Weighted Currencies in the
Dallas Index (in percent)
Country

1976

1985

1995

Brazil
Canada*
China
France*
Germany*
Italy*
Japan*
Korea
Mexico
Netherlands*
Saudi Arabia
Singapore
Taiwan
United Kingdom*
Venezuela
Total weight of top 10

2.3
22.2
†
2.7
5.9
2.8
11.7
†
4
2.7
†
†
†
4.5
2.9
61.7

†
19.4
†
2.7
4.8
†
14.3
2.8
5.7
2.3
2.6
†
3.5
5.1
†
63.2

†
20.3
3.9
2.8
4.7
†
15
3.1
8.1
†
†
2.3
3.9
4.5
†
68.7

* Country whose currency is included in the Board index.
† Not in the top 10 in this year.

nually using a three-year moving average.
The CmupBoard index therefore is identical to the Dallas index except that it includes only 12 currencies, not 128.

FEDERAL RESERVE BANK

13

OF

S T. L O U I S

J U LY /A U G U S T 1996

yen is more than seven percentage points
higher in the BilBoard index than in the
Board index, whereas the other eight
countries receive less weight in the BilBoard index than in the Board index.
These weight changes produce some
noteworthy differences in the two indexes
that are shown in the top panel of Figure
3. Table 3 reveals that between January
1976 and March 1985, the dollar appreciated 43 percent according to the Board
index and 30 percent according to the BilBoard index. Accounting for this difference is relatively straightforward. The U.S.
dollar rose by less against the Canadian
dollar during the 1976–85 period than it
did against some currencies that received
higher weights than the Canadian dollar in
the Board index (for example, the French
franc and the British pound). With respect
to the Japanese yen, the U.S. dollar fell
during the 1976–85 period. Furthermore,
since March 1985, the dollar has changed
little relative to the Canadian dollar, falling
only 1 percent. The dollar has fallen far
more against the remaining nine currencies since 1985. As a result, the BilBoard
index shows a less pronounced change in
the dollar over the sample period than
does the Board index.
The direction of the movement in the
BilBoard index, however, closely matches
that of the Board index as shown by the
high degree of correlation between the two
in Table 4. The correlation was 0.98 over
the entire period. Meanwhile, the correlation between the BilBoard index and the
Dallas index, even though high during
1976–85, is negative during 1985–95 and
negative over the entire sample period
1976–95. In sum, the differences between
the Board and the Dallas indexes cannot
be primarily attributed to a difference in
the method used to calculate the weights
of each currency.

Figure 3

Constructed Effective Exchange Rates

Index

Index

Index

(March 1985=100)
200
180
160
140
120
100
80
60
40
20
200
180
160
140
120
100
80
60
40
20
200
180
160
140
120
100
80
60
40
20

Board Dallas BilBoard BupBoard

76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
Year
Board Dallas MupBoard

76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
Year
Board Dallas CmBoard CmupBoard

76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
Year

Bilateral vs. Multilateral Trade
Shares—BilBoard
As shown in Table 7, the weights assigned to each currency in the Board and
the BilBoard indexes vary substantially.
For example, the weight given to the
Canadian dollar is more than 30 percentage points higher in the BilBoard index
than in the Board index. The reason for
this difference is that although Canada is
the most important U.S. trading partner, it
is less important in worldwide trade. Japan
also holds a higher share of U.S. trade than
worldwide trade, but the other eight countries rank higher in worldwide trade rather
than in trade with the United States. As a
result, the weight given to the Japanese

Base Period for Trade
Weights—MupBoard
The multilateral trade shares of the
countries used in the MupBoard index for
1976, 1985, and 1994 are shown in Table 7.

FEDERAL RESERVE BANK

14

OF

S T. L O U I S

J U LY /A U G U S T 1996

Table 7

Trade Weights for Constructed Indexes (percent)
MupBoard
Country
Belgium
Canada
China
France
Germany
Italy
Japan
Mexico
Netherlands
Sweden
Switzerland
United Kingdom

BupBoard

CmupBoard

Board

BilBoard

1976

1985

1994

1976

1985

1995

CmBoard

1976

1985

1995

6.4
9.1
0
13.1
20.8
9.0
13.6
0
8.3
4.2
3.6
11.9

3.4
39.9
0
4.8
10.1
4.8
21.0
0
4.6
1.6
1.9
7.9

6.4
9.0
0
12.7
20.6
9.1
13.6
0
8.1
4.2
4.6
11.9

5.9
8.9
0
12.1
19.1
9.6
17.0
0
7.2
3.4
4.7
12.0

6.3
7.8
0
12.4
21.7
9.9
15.8
0
7.0
2.9
4.5
11.5

3.5
39.3
0
4.8
10.4
4.9
20.8
0
4.8
1.6
1.9
8.0

3
35.6
0
5.0
8.8
4.1
26.3
0
4.3
1.6
2.0
9.4

2.8
37.3
0
5.1
8.6
3.7
27.5
0
3.3
1.3
2.0
8.3

3.1
37.2
0.5
4.4
9.4
4.5
19.5
6.3
4.3
1.5
1.8
7.4

3.2
36.5
0.6
4.4
9.6
4.6
19.3
6.6
4.4
1.5
1.8
7.4

2.7
31.6
1.8
4.4
7.8
3.6
23.3
9.3
3.8
1.4
1.8
8.3

2.3
30.5
5.9
4.2
7.1
3.0
22.6
12.2
2.7
1.1
1.7
6.8

* Note that weights in the Board index are based on multilateral trade shares during 1972–76. Weights in the BilBoard and CmBoard indexes are based on bilateral trade shares during 1972-76. Weights in the
MupBoard, BupBoard, and CmupBoard indexes are based on three-year moving average bilateral trade shares, updated annually. Thus, the weights in the three columns: 1976, 1985, and 1995 (1994 for
MupBoard), are based on trade shares during 1973–75, 1982–84, and 1992–94, (1991–93 for MupBoard), respectively.

These trade shares did not change substantially over time. As a result, the MupBoard
index closely mimics the Board index, as
shown in the middle panel of Figure 3. Both
indexes show the same percentage appreciation of the dollar between January 1976 and
March 1985 and nearly the same depreciation from March 1985 through 1994.22 Likewise, the two indexes were nearly perfectly
correlated. Thus one can conclude that the
frequency of updating weights is not the driving force for differences in the Board and
Dallas indexes.

the other currencies included in the
index.
The behavior of the BupBoard index
resembles that of the Board index. For example, Table 3 shows a 17 percent depreciation of the dollar using the BupBoard
index from January 1976 to December
1995, whereas the Board index shows a
19 percent depreciation of the dollar. During this period the Dallas index shows the
dollar appreciating by 137 percent. These
results are reinforced by the correlation
coefficients shown in Table 4. The BupBoard index is highly correlated with the
Board index in the 1976–95 period (0.97)
but negatively correlated with the Dallas
index (20.45). Changing the manner and
frequency with which the weights are calculated to accord with the Dallas index did
not create an index that resembled the Dallas index. Thus the primary cause of the
differences between the two indexes must
be the selection of countries in each index.
The CmBoard index allows us to further explore the importance of country
choice. In the CmBoard index, the weights
given to each currency are determined by
that country’s share of trade with the

Currency Choice—BupBoard,
CmBoard and CmupBoard
The top panel of Figure 3 shows that
the BupBoard index closely mimics the behavior of the BilBoard index, particularly
in the 1976–85 period when the weights
for the two indexes, listed in Table 7, are
similar. In the 1985–95 period, as Japan’s
share of U.S. trade rises, the BupBoard
index shows a slightly larger depreciation of the dollar than the BilBoard index.
This result follows from the fact that
during this period the U.S. dollar fell by
more against the yen than against any of

FEDERAL RESERVE BANK

15

OF

S T. L O U I S

22

Worldwide trade data for some
of the countries used in the
index were not available for
1994; therefore, the MupBoard index ends in 1994.

J U LY /A U G U S T 1996

23

We were unable to construct
an index using multilateral
trade shares that included
China and Mexico because
world trade data for China
before 1982 are unavailable.

United States.23 This index therefore differs
from the Board index in two ways: its inclusion of China and Mexico and the use
of bilateral trade shares. The behavior of
the CmBoard index, shown in the bottom
panel of Figure 3, is similar to the Board
index over the January 1976–March 1985
period. As shown in Table 3, the CmBoard
index appreciated 46 percent, whereas the
Board index appreciated 43 percent. A
greater difference between the CmBoard
and the Board indexes occurs over the period from March 1985 to December 1995.
The CmBoard index shows an 18 percent
trade-weighted depreciation of the dollar
during this period, while the Board index
shows a 62 percent depreciation. The CmBoard index, however, does not show an
appreciation of the dollar as the Dallas
index does during this period. That the
changes embedded in the CmBoard index
cause it to become more similar to the
Dallas index and less similar to the Board
index is reinforced by the correlation coefficients in Table 4. For the entire period,
the correlation of the CmBoard index with
the Board index is much lower than the
Bilboard, MupBoard, and BupBoard indexes, whereas its correlation with
the Dallas index is positive rather than
negative.
The CmupBoard index, which also includes China and Mexico, still does not
show the magnitude of the appreciation of
the dollar in the bottom panel of Figure 3
that the Dallas index indicates in the January 1976–March 1985 period. In contrast,
however, to all of the previously constructed indexes, it does show an appreciation of the dollar during the March
1985–December 1995 period, although
this appreciation is less than that indicated
by the Dallas index. For the entire period,
the CmupBoard index shows little correlation with the Board index but is highly
correlated with the Dallas index.
The CmBoard and the CmupBoard indexes illustrate two key points. The first is
that the Dallas index differs from the
Board index primarily because the Dallas
index includes currencies whose behavior,
particularly during the March 1985–

December 1995 period, was in sharp contrast to the behavior of the currencies included in the Board index. Specifically, the
Dallas index includes currencies against
which the dollar appreciated substantially
during this period. Between March 1985
and December 1995, the dollar rose by
362 percent against the Mexican peso. In
contrast, the dollar fell against all of the
currencies included in the Board index
during this period.
The second point is that in an index in
which there are sharp differences in the
behavior of the currencies (such as the
Dallas index), the weights assigned to each
currency matter. In the Board index the
behavior of the currencies was relatively
similar: The dollar rose against all 10 currencies with the exception of the Japanese
yen during the early period and fell
against all 10 currencies during the later
period. Given such similarities in the behavior of the currencies, the manner in
which the weights were calculated—bilateral or multilateral trade shares—and the
frequency of updating of the weights had
little effect on the behavior of the indexes.
However, when the behaviors of the currencies in the index differ greatly, as evidenced by the enormous appreciation
of the dollar against the Mexican peso
during the same period in which the
dollar was depreciating against the currencies of the major industrialized countries,
the method of calculating the weights
assigned to each currency increases in importance.
This latter point is illustrated by the
differences in the CmBoard and the CmupBoard index. The dollar appreciated
against the Chinese yuan by 107 percent
between March 1985 and December 1995.
This appreciation, however, has little effect
on the trade-weighted value of the dollar
when the weight assigned to the yuan is
based on China’s share of U.S. trade over
the 1972–76 period (as in the CmBoard
index). With annual updates of the
weights, as in the CmupBoard index, the
growth in China’s share of U.S. trade
places increased importance on the appreciation of the dollar against the yuan.

FEDERAL RESERVE BANK

16

OF

S T. L O U I S

J U LY /A U G U S T 1996

Likewise, the appreciation of the dollar
against the peso is given greater weight in
the index with annual updates. If the
weights used in the CmBoard index had
been based on the 1992–94 trade shares,
the index would have shown a sharper appreciation of the dollar than that evidenced by the CmupBoard index.
The difference between the Board and
the Dallas indexes does not simply result
from the fact that the Dallas index includes more countries than the Board
index. Two factors make the country
choice important: (1) the Board index excludes (the Dallas index includes) countries that account for a significant share of
U.S. total merchandise trade; and (2) the
behavior of the excluded currencies
against the dollar has been substantially
different since 1985 from that of the currencies included in the Board’s index. The
importance of the first factor has increased
over time. In 1976, as shown in Table 6,
seven of the 10 currencies that constitute
the Board index were among the 10 most
heavily weighted currencies in the Dallas
index. By 1995, only five of the countries
included in the Board index also were in
the top 10 of the Dallas index.
Our analysis indirectly identifies an
important consideration in using tradeweighted exchange rate indexes as a measure of international competitiveness.
Generally speaking, changes in real (that
is, nominal exchange rates adjusted for inflations difference), rather than nominal
exchange rates, are commonly used for assessing changes in international competitiveness. Since the inflation experience of
the countries whose currencies are in the
Board index has been roughly similar over
time, the nominal Board index mimics its
real counterpart. The Dallas index, however, includes countries that have experienced periods of hyperinflation. As a result
of this hyperinflation, the currencies of
these countries depreciated sharply against
the dollar during these periods, driving the
appreciation of this index between 1985
and 1995. After adjusting for the inflation
differences, the real Dallas index declines
between 1985 and 1995.

CONCLUSION
Our examination of effective exchange
rates reveals the many decisions underlying their construction. These decisions can
produce substantially different views of
changes in the average foreign exchange
value of a currency. The actual effect of
these decisions was investigated by comparing the Board index with the Dallas
index.
The difference between the Board
index and the Dallas index is driven primarily by the choice of currencies. This
does not mean, however, that issues such
as the determination of trade shares and
the frequency with which weights are updated are unimportant. What makes these
latter factors unimportant in the Board
index is the similarity in the behavior of
the currencies that make up the index.
This also illustrates why all of the tradeweighted exchange rate indexes covered in
this article show an appreciation of the
dollar between 1976 and 1985. During
this period, and particularly after 1980,
the dollar was appreciating against most
other currencies. Since 1985, the behavior
of the dollar has been markedly different
against the currencies of the industrialized
countries from its behavior against the
currencies of the developing countries.
Thus even though we have not provided a
definitive answer to the question posed in
the title of this article, the reasons for the
measurement differences have been illuminated.

REFERENCES
Batten, Dallas S., and Michael T. Belongia. “Do the New Exchange Rate
Indexes Offer Better Answers to Old Questions?” this Review (May
1987), pp. 5–17.
Belongia, Michael T. “Estimating Exchange Rate Effects on Exports:
A Cautionary Note,” this Review (January 1986), pp. 5–16.
Board of Governors. “Index of the Weighted-Average Exchange Value of
the U.S. Dollar: Revision,” Federal Reserve Bulletin (August 1978),
p. 700.
Coughlin, Cletus C.; Patricia S. Pollard; and Jerram C. Betts. “To
Chain or Not to Chain Trade-Weighted Exchange Rate Indexes,”
Federal Reserve Bank of St. Louis Working Paper No. 96-010A
(1996).

FEDERAL RESERVE BANK

17

OF

S T. L O U I S

J U LY /A U G U S T 1996

Cox, W. Michael. “A New Alternative Trade-Weighted Dollar Exchange
Rate Index,” Federal Reserve Bank of Dallas Economic Review
(September 1986), pp. 20–8.
Deephouse, David L. “Using a Trade-Weighted Currency Index,” Federal
Reserve Bank of Atlanta Economic Review (June/July 1985), pp.
36–41.
Hargreaves, Derek. “Effective Exchange Rates: OECD Currencies,” Morgan Guaranty Trust Company Economic Research Note (December 30,
1993).
Hirsch, Fred, and Ilse Higgins. “An Indicator of Effective Exchange
Rates,” IMF Staff Papers (November 1970), pp. 453–87.
Hooper, Peter, and John Morton. “Summary Measures of the Dollar’s
Foreign Exchange Value,” Federal Reserve Bulletin (October 1978),
pp. 783–9.
Ott, Mack. “The Dollar’s Effective Exchange Rate: Assessing the Impact
of Alternative Weighing Schemes,” this Review (February 1987),
pp. 5–14.
Rhomberg, Rudolf R. “Indices of Effective Exchange Rates,” IMF Staff
Papers (March 1976), pp. 88–112.
Rosensweig, Jeffrey A. “Constructing and Using Exchange Rate Indexes,” Federal Reserve Bank of Atlanta Economic Review (Summer
1987), pp. 4–16.
______. “A New Dollar Index: Capturing a More Global Perspective,”
Federal Reserve Bank of Atlanta Economic Review (June/July
1986a), pp. 12–22.
______. “The Atlanta Fed Dollar Index and its Component SubIndexes,” Federal Reserve Bank of Atlanta Working Paper No. 86-7
(August 1986b).
Turner, Philip and Jozef Van ‘t dack. “Measuring International Price and
Cost Competitiveness,” BIS Economic Papers (November 1993).

FEDERAL RESERVE BANK

18

OF

S T. L O U I S

J U LY /A U G U S T 1996

Michael R. Pakko is an economist at the Federal Reserve Bank of St. Louis. David C. Wheelock is a research officer at the Federal Reserve
Bank of St. Louis. Heidi Beyer and Kelly Morris provided research assistance.

Clearly, the statement that the Fed
controls interest rates is, at best, an oversimplification. This article attempts to
demystify the relationship between Federal
Reserve monetary policy actions and
interest rate behavior. Interest rates are set
in competitive markets by factors affecting
the supply of and demand for individual
securities. Monetary policy actions can
affect both the supply of and the demand
for financial assets, and their effects
depend not only on current actions but
also on the public’s expectations of future
policy moves.
We describe in some detail the nearterm behavior of government security
yields following three recent Federal Reserve policy actions. On the most recent
occasion, the Fed’s easing action on
January 31, 1996, market yields changed
little immediately following the policy
move, but then yields rose over succeeding
months. We contrast this experience with
two other events. In early 1994, Fed policy
moves to raise interest rates were associated with increases in market interest
rates that might be considered greater than
justified by the extent of Fed actions.
Then, in May 1994, market yields declined
following a Fed policy action that was
widely interpreted as an effort to raise
interest rates. Our review of these episodes
reveals how expectations of future monetary policy actions, expectations of the
effect of policy on future inflation, as well
as nonmonetary influences can cause
market interest rates to behave in diverse
ways after apparently similar Fed actions.
We begin with a brief description of
how the Fed carries out open market
policy and the channels through which
policy might affect market interest rates.
Next, we examine some recent episodes in
which market interest rates responded in
different ways to Federal Reserve policy
moves. Finally, we conclude with a
summary of how perceptions of future
monetary policy actions affect the behavior

Monetary Policy
and Financial
Market
Expectations:
What Did They
Know and
When Did They
Know It?
Michael R. Pakko and
David C. Wheelock

O

n January 31, 1996, the Federal
Open Market Committee (FOMC)
voted to ease monetary policy, which
was widely reported as a lowering of interest
rates. Although some interest rates fell
with the Fed’s action, the declines were
generally small, and over succeeding
months market interest rates tended to rise.
The yield on the Treasury’s 10-year note,
for example, which had been 5.63 percent
on January 30, and which closed at
5.60 percent on January 31, stood at
6.34 percent on March 29, and reached
7.03 percent by June 12. Other rates
behaved similarly over this period.
Such seemingly perverse moves in
market interest rates have also followed
other monetary policy actions, sometimes
even on the day those actions were taken.
Commonly, Federal Reserve moves to raise
or to lower interest rates are followed by
changes in market interest rates in the
same direction. On May 17, 1994, however, the Fed announced a tightening of
monetary policy, which some might expect
would cause market interest rates to rise.
Instead, many market rates immediately
declined.

FEDERAL RESERVE BANK

19

OF

S T. L O U I S

J U LY /A U G U S T 1996

Figure 1

Federal Funds Rate Target
(1984-1996)
12
11
10
9
8
7
6
5
4
3
2

Percent

Percent

9
8
7
6
5
4
3
2
84

85

86

87

88

89

90

91

92

93

94

95

96

federal funds rate. The Fed can have a considerable effect on the federal funds rate
because its
open Rate
marketTarget
operations affect
Federal
Funds
the
aggregate
supply
of
bank
reserves.
(1984-1996)
It is generally acknowledged that the
Fed has considerable influence on the
equilibrium federal funds rate, at least for
relatively short periods. But do Federal
Reserve operations affect other market
interest rates?

The Expectations Hypothesis
The expectations hypothesis of interest
rate determination states that long-term
interest rates will reflect current and
expected future yields on short-term securities. For example, the yield on two-year
Treasury notes should be the average of
the current yield on one-year Treasury bills
and the expected yield on one-year bills
whose holding period begins one year
from now. Interest rate arbitrage ensures
that this will occur. If, for example, the
interest rate on one year securities that is
expected to prevail one year from now
would suddenly decrease, arbitrage would
cause the current demand for two-year
securities to rise. This would tend to lower
the market yield on two-year securities to
an average of the current one-year yield
and the (now lower) one-year yield
expected to prevail one year from now.
Similarly, the yield on three-month
Treasury bills should reflect the current
and expected future path of the federal
funds rate over the next three months. As
a result, changes in current or expected
future short-term interest rates will tend to
cause similar movement all along the yield
curve.2
Because long-term rates are linked to
the current and expected future path of
short-term interest rates, expectations of
future Fed policy moves are important to
the movements of interest rates today. It is
significant therefore that changes in the
Fed’s target for the federal funds rate tend
to be persistent, with a series of changes
accumulating over time. This tendency is
clearly illustrated in Figure 1, which
shows how the Fed’s target has evolved
over the past several years.

97

of market interest rates in response to current policy moves and hence complicate
the assessment of the Fed’s credibility as an
inflation fighter.

MONETARY POLICY,
EXPECTATIONS AND
MARKET INTEREST RATES

Open Market Operations and
Short-Term Interest Rates

1

The Fed also sets the discount
rate, which is the rate charged
banks when they borrow
reserves from the Fed, and
required reserve ratios, that is,
the percentage of their deposit
liabilities that banks are
required to hold in the form of
vault cash or deposits at
Federal Reserve Banks. Neither
is changed frequently, however,
and open market policy is the
principal mechanism by which
the Fed conducts monetary
policy.

2

See Campbell (1995) for more
detail about the term structure
of interest rates and empirical
evidence on the expectations
hypothesis.

Although Federal Reserve monetary
policy is often described in press accounts
as the manipulation of interest rates, in
fact, monetary policy is carried out mainly
by varying the supply of reserves available
to the banking system.1 Open market purchases of Treasury securities by the Fed
supply additional reserves, whereas open
market sales withdraw reserves.
Banks hold reserves to meet statutory
requirements, as well as to meet the payment demands of their customers. A bank
with a reserve deficiency might borrow
reserves from the Fed, sell securities from
its portfolio, or borrow reserves by purchasing federal funds in the interbank
reserves market. Similarly, banks with surplus reserves may choose to convert their
surpluses into earning assets by acquiring
securities or other assets or by selling federal funds. The interest rate that clears the
market for federal funds is known as the

FEDERAL RESERVE BANK

20

OF

S T. L O U I S

J U LY /A U G U S T 1996

pay this inflation premium for the same
reason. As a result, a fundamental relationship between inflation and interest rates is
given by the Fisher relationship,

Financial market participants are well
aware of this pattern. For example, after an
increase in the federal funds rate target on
February 4, 1994, the Wall Street Journal
reported, “There is little disagreement on
where short-term interest rates will be
going over the next year: up. The only
question is how far they will rise and how
fast.”3
The persistence in federal funds rate
changes causes current movements of the
funds rate to provide information about
future changes. When evaluating the
course of short-term interest rates over
several months, a current increase
(decrease) can be expected to result in further increases (decreases). Because
longer-term interest rates are affected by
anticipated changes in short-term rates,
the yield on a given security might
respond to a particular change in the federal funds rate by more than the amount of
the funds rate change because the security
yield will incorporate the expectation of
future changes in the funds rate in the
same direction.

i = r + πe,

(1)

which states that the nominal interest rate
(in dollar terms) consists of the following
two components: the real interest rate (r)
and a component that equals expected
inflation (πe).
Thus if market participants interpret a
monetary policy action as providing new
information about the outlook for inflation, interest rates should change accordingly. This is referred to as the expected
inflation effect of monetary policy on
interest rates. Financial market participants
who are interested in the future course of
inflation watch Federal Reserve actions
closely. If the Fed is viewed as likely to
pursue a policy that will prevent
significant inflation over time, market
yields will be lower. On the other hand,
if the public doubts that the Fed is committed to low inflation, then financial
markets will reflect fears of future inflation
by incorporating an inflation premium in
interest rates.
When investors are uncertain about
the future course of monetary policy, and
hence are uncertain about the future
course of inflation, market yields might
also be higher than they otherwise would
be. For example, although inflation fell
substantially in the early 1980s, interest
rates remained high, and the difference
between the level of market interest rates
and the concurrent inflation rate has only
recently declined to approximate the
difference of the early 1960s. In other
words, the ex post real interest rate—
the difference between the market, or
nominal, interest rate and the rate of inflation—was unusually high (see Figure 2).
One interpretation of the high ex post
real interest rates of the 1980s is that, after
experiencing rising inflation from 1965 to
1979, investors feared a return of high
inflation and thereby demanded high nominal returns on fixed-income assets.

Monetary Policy, Inflation
Expectations, and the Fisher
Relationship
Interest rate arbitrage can explain why
market interest rates often move upward
when the Fed raises its federal funds
target, and downward when the Fed
lowers its target. Sometimes, however,
market rates fall when the Fed raises its
target and rise when the Fed lowers its
target. Such apparently perverse changes
in market rates can occur because Fed
operations are not the sole influences on
the supply of and demand for securities.
Such changes can also happen because
monetary policy is the principal determinant of the long-run rate of inflation—
and inflation can have a pronounced effect
on interest rates.
Because inflation erodes the purchasing power of money, an increase in inflation causes lenders to require higher
interest rates as compensation for receiving future payments in money that has
declined in value. Borrowers are willing to

FEDERAL RESERVE BANK

21

OF

S T. L O U I S

3

Thomas T. Vogel, Wall Street
Journal, February 7, 1994,
p. C1.

J U LY /A U G U S T 1996

commercial bank reserve positions,” which
it anticipated would increase market
interest rates (specifically, the Fed had
increased its objective for the federal
funds rate by 25 basis points to 3.25
percent).
The official announcement of such a
move was unprecedented, and the FOMC
stated that it had made the announcement
in part because this was the first tightening
of monetary policy since 1989.5 Although
it was noted that such a public announcement should not be interpreted as precedent
setting, after its meeting on February 2,
1995, the FOMC announced that after
each future meeting it would issue a statement indicating whether there would be
any change in policy.
By publicly announcing specific policy
moves, the FOMC has eliminated uncertainty about its current operational stance.6
But because the future course of policy
remains uncertain, market participants
continue to expend considerable effort
attempting to forecast upcoming policy
actions. Speculation about possible nearterm actions often seems to affect the
market prices and trading volumes of
financial assets as much as actual
moves do.

Figure 2

Ten Year Government Security Yield and
Year-Over-Year CPI Inflation
(January 1959-June 1996)
16 Percent

Security Yield

14
12

Percent 16
14
12

Inflation Rate

10

10

8

8

6

6

4

4

2

2
0

0
59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95

4

See Dotsey and DeVaro
(1995) for empirical evidence
suggesting that much of the
disinflation of the early 1980s
was unanticipated by the
public.

5

See Pakko (1995) for a
detailed description of FOMC
policy moves during 1994 and
Gavin (1996) for a discussion
of policy moves during 1995.

6

Thornton (1996) finds that
financial market volatility has
been lower around the time of
FOMC meeting dates since the
policy of announcing federal
funds rate changes was implemented.

Alternately, if investors viewed the prospects for economic growth favorably, they
may have simply demanded higher real
returns on fixed-income investments. Still,
because disinflations are often accompanied by high ex post real rates, both in
the United States and in other countries
[see, for example, Dueker (1996)], a fear
of renewed inflation is a plausible explanation for high real rates in the 1980s.4
Carlstrom (1995) has aptly referred to
this effect of Federal Reserve policy on
interest rates as a monetary policy paradox. Short-term interest rates can be
lowered only by increasing monetary
growth, which tends to kindle inflationary
expectations and higher interest rates.
Lowering interest rates in the long
run may require raising them in the
short run.

Expectations and Treasury Security
Yields
Figure 3 plots the market yields on
three U.S. Treasury securities on the date
of each announced change in open market
policy, that is, change in intended federal
funds rate, and each meeting of the FOMC
during 1994, 1995, and January 1996. The
Fed increased its federal funds target six
times in 1994 and once in 1995; the Fed
reduced its target twice in 1995 and again
on January 31, 1996. The change in basis
points, if any, in the Fed’s target is noted
near the top of each vertical line corresponding to the date of a policy change or
FOMC meeting. The market yields on
three-month Treasury bills, one-year Treasury bills, and 10-year Treasury notes on
each date are plotted, as are the yields five
business days before and five business days
after the central dates.

MONETARY POLICY AND
INTEREST RATES IN THE
SHORT RUN
To evaluate the effect of Federal
Reserve policy actions, we focus on the
behavior of market interest rates on dates
immediately preceding and immediately
following recent actions. The Fed made no
changes in its target for the federal funds
rate during 1993, but on February 4, 1994,
the FOMC announced that it had voted to
“increase slightly the degree of pressure on

FEDERAL RESERVE BANK

22

OF

S T. L O U I S

J U LY /A U G U S T 1996

Figure 3

The Market Response to Changes in the Fed Funds Target
and FOMC Meetings
10
9

Three Month

Interest Rates
+25

+25

+25

+50

0

+50

2/4/94

3/22/94

4/18/94

5/17/94

7/6/94

8/16/94

0

One Year
+75

Ten Year
0

8
7
6
5
4
3
2

9

9/27/94

11/15/94

12/20/94

-25

-25

12/19/95

1/31/96

Interest Rates
+50

0

0

-25

0

0

5/23/95

7/6/95

8/22/95

9/26/95

0

8
7
6
5
4
3
2
2/1/95

3/28/95

Market yields tended to rise during
1994, coincident with the Fed’s target rate
increases. Yields generally fell in 1995, and
the differences in yields of securities with
different maturities narrowed. The term
structure of yields is often interpreted as
revealing market expectations about the
future paths of real returns and inflation.
Researchers—including Fama (1990),
Mishkin (1990), and Estrella and Mishkin
(1995)—conclude that yield spreads contain both types of information. Long-term
rates tend to be sensitive to inflation

11/15/95

expectations, whereas short-term rates
follow current and expected real shortterm rates more closely. Hence the substantial narrowing in the yield spread
across securities of different maturities
during 1995 could reflect diminished
expectations for real returns, inflation,
or both.
On February 1, 1995, the Fed made
the last in a series of federal funds target
increases. Although market interest rates
rose that day, on subsequent days they
resumed a decline that had begun in late

FEDERAL RESERVE BANK

23

OF

S T. L O U I S

J U LY /A U G U S T 1996

THE FEDERAL FUNDS FUTURES MARKET

†

Federal funds futures (formally known as 30-Day Interest Rate futures) have been
actively traded at the Chicago Board of Trade since October 1988. The federal funds
futures contract is based on the monthly average federal funds rate as reported by the
Federal Reserve Bank of New York.
The contract itself calls for delivery of the interest paid on a principal amount of
$5 million in overnight federal funds held for 30 days. Contracts are priced in units
of 100, with the federal funds rate being 100 minus the price (for example, a price of
92.75 implies a 7.25 percent funds rate). Contracts are settled daily, with the purchaser of a contract paying the seller $41.67 (per $5 million contract) for each basis point
increase in the implied federal funds rate (or each 1/100 of a point decline in the contract price) at the close of business. This tick size has been set by using a 30-day
month: $5 million 3 30/360 3 0.0001 = $41.67.
The following example helps explain the potential hedging use of federal funds
futures. Consider a bank that is a consistent buyer of $75 million in federal funds at a
current rate of 7 percent. The bank is worried that the federal funds rate will rise in
the current month, raising its cost of funds. By selling 15 futures contracts (15 3 $5
million = $75 million), the bank stands to profit from the futures transactions in the
event that it suffers a loss from a higher cost of funds. For instance, suppose that on
the first day of the month, the bank purchases the contracts at 93.00—implying a federal funds rate of 7 percent. If the funds rate immediately rises to 7.2 percent, the
bank ends up paying $450,000 in interest on its federal funds purchases over the
course of the month [$75 million 3 .0720 3 (30/360)]. However, the buyer of the
federal funds futures contract pays the bank $12,501 [15 contracts 3 20 ticks 3
$41.67]. The net cost to the bank is $437,499. The bank’s effective cost of funds has
been locked in at 7 percent [($437,499/$75 million) 3 (360/30)].
In addition to banks like the one described in the preceding example—seeking to
hedge positions in the federal funds market—futures trade is also carried out by speculators who are betting on a particular course of monetary policy. Each type of trader
has an incentive to consider the most likely outcome of monetary policy when deciding whether to participate in a transaction, so the price of federal funds futures represents the market’s best estimate of the federal funds rate over the course of the
contract month.
†

A more complete description of the federal funds futures market can be found in Chicago Board of Trade (1995).

1994. Security yields continued to decline
throughout 1995, with the Fed lowering
its funds rate target in July and December
and again in January 1996.
It is apparent from Figure 3 that when
the Fed changes its federal funds target,
market rates sometimes, but not always,
move in the same direction as the Fed’s
adjustment. Even when market rates do
move in the same direction, they do not
move by the same amount as the change
in the federal funds rate. A change in
expected inflation accompanying a monetary policy action could explain otherwise
counterintuitive changes in market interest

rates, such as a decline in market rates following a tightening of monetary policy or
an increase in market rates following an
easing of policy.
In the next sections we examine in
more detail the behavior of market rates
around three recent episodes of changes in
the Fed’s target federal funds rate. Knowledge of the extent to which financial
market participants anticipated a policy
move is important for interpreting each
event. Monetary policy actions that are
widely anticipated will not convey new
information about future inflation, but
actions that take markets by surprise may

FEDERAL RESERVE BANK

24

OF

S T. L O U I S

J U LY /A U G U S T 1996

alter forecasts of future inflation. The
effect of a policy move on interest rates
thus depends on whether the move was
expected. One source of information about
market expectations of Fed policy moves is
the federal funds futures market.

Figure 4

Federal Funds Futures
Actual, 1 month forward and 2 months forward through 3/96
12

Information from the Federal Funds
Futures Market

Percent

Percent

Actual
1 Month Forward
2 Months Forward

10

Since 1988, the Chicago Board of
Trade has offered a market in futures contracts based on the federal funds rate.
(See the shaded box, The Federal Funds
Futures Market.) Contracts in this market
are based on the monthly average federal
funds rate, as reported by the Federal
Reserve Bank of New York. The market
is used both by financial institutions to
hedge their federal funds market positions
against changes in the funds rate and by
speculators attempting to predict Federal
Reserve monetary policy. Because the
contracts are based on future monthly
averages of the federal funds rate, price
movements directly reflect market participants’ expectations of policy actions.
Figure 4 shows the accuracy with
which the federal funds futures market has
predicted actual movements in the funds
rate. Both the one-month ahead and twomonth ahead rates track the actual funds
rate closely, although the two-month ahead
forecast fails to predict turning points
as accurately as the one-month ahead
forecast, lagging behind actual funds rate
movements. Nevertheless, Krueger and Kuttner (1995) and Rudebusch (1996) find that
one-month, two-month, and three-month
future rates are all accurate predictors of
subsequent federal funds rate movements.
Information from the federal funds
futures market is used in Figure 5 to show
expectations of movements in the funds
rate implied by futures prices in the days
leading up to and following FOMC meetings and policy changes in 1994 and 1995.
The figure shows two series of futures
yields. One series is the funds rate the
market predicts will prevail after the meeting (see the appendix for details of the
calculations). The second series is the
funds rate derived from a three-month for-

10

8

8

6

6

4

4

2

2
88

89

90

91

92

93

94

ward contract, indicating market expectations for future levels of the federal
funds rate.
Figure 5 illustrates several notable
points. First, the three-month ahead
futures rate was above the one-month
futures rate throughout 1994 and into
early 1995. But when the Fed lowered the
funds rate in July 1995, its first such move
since 1992, the three-month futures rate
was below both the spot rate and the current month’s predicted funds rate. The
market had thus correctly forecast the
directional change in Fed policy.
The data in Figure 5 also show that
many of the Fed’s policy actions during
1994 were at least partly anticipated. That
is, futures contracts were priced to reflect
changes in the federal funds rate before the
Fed altered its target. On occasions when it
appears that funds rate changes were not
fully anticipated, the three-month forward
forecast moved in the same direction as
the forecast funds rate for the remainder
of the current month. In other words,
unexpected changes in the Fed’s target
led market participants to expect further
adjustments to the rate in the same
direction as the initial move. The evidence
thus indicates that, at least since 1994,
the federal funds futures market has
forecast specific Fed policy actions fairly
well and that futures prices reflect

FEDERAL RESERVE BANK

25

OF

12

S T. L O U I S

95

96

J U LY /A U G U S T 1996

Figure 5

Fed Funds Expectations Derived from
7.5
7.0

Percent
+25

+25

+25

+50

+0

+50

5/17/94

7/6/94

+0

+75

+0

6.5
6.0
5.5
5.0
4.5
4.0
3.5
3.0

2/4/94

3/22/94

4/18/94

8/16/94

9/27/94 11/15/94 12/20/94

Percent
7.5
+50

+0

+0

-25

0

0

7/6/95

8/22/95

0

-25

-25

7.0
6.5
6.0
5.5
5.0
4.5

2/1/95

3/28/95 5/23/95

Expected Fed Funds

3-Month Futures

9/26/95 11/15/95 12/19/95 1/31/96

Target Before Meeting

the Fed’s tendency to make multiple
moves in one direction before reversing
course.

Target After Meeting

interest rates would be useful. As our
analysis suggests, however, the effects of
monetary policy moves on interest rates can
be difficult to disen-tangle. This difficulty is
illustrated by a look at three specific
episodes of Federal Reserve policy moves.

Evaluating Market Responses to
Specific Monetary Policy Actions
For both policymakers and market participants, the information about expected
monetary policy and inflation embedded in

February 1994

On February 4, 1994, the FOMC voted

FEDERAL RESERVE BANK

26

OF

S T. L O U I S

J U LY /A U G U S T 1996

to “increase slightly the degree of pressure
on reserve positions,” resulting in an
increase of 25 basis points in the federal
funds rate. At the time, some financial analysts claimed that the move took markets by
surprise. The move, however, was foreshadowed by Federal Reserve Chairman Alan
Greenspan only days earlier. On January 31,
1994, the chairman stated that, “at some
point . . . we will need to move [short-term
interest rates] to a more neutral stance.”7
This comment was interpreted by some
analysts as indicating, “It’s a question of
when, not whether, they will tighten.”8
The path of federal funds rate expectations illustrated in Figure 5 makes it clear
that the move was anticipated. Beginning
on January 31, the expected funds rate
rose gradually to the point where the
25 basis point move was almost fully anticipated on the day it occurred. Figure 3
shows that long-term interest rates rose
along with the expected federal funds rate.
However, bond rates tended to rise by
more than the expected funds rate. From
January 28 through February 4, the
expected federal funds rate rose by 22
basis points, whereas the three-month,
one-year, and 10-year Treasury security
yields rose by 30, 35, and 26 basis points,
respectively.
There are many potential explanations
for the larger increases in Treasury security
yields. One explanation is rather unique to
this particular occasion. It holds that the
Fed’s policy adjustment was a preemptive
move to head off a possible rise in inflation
rather than a response to an alreadyobserved increase in inflation. Yet many
observers had not seen the emergence of
inflation as imminent, so the move was
interpreted by some as indicating that the
FOMC had information or insight about
inflation that was not generally available
to the public. Hence inflation expectations
were revised upward, and market yields
rose.
A related explanation for the large
increases in security yields is that the
public viewed the relatively small policy
move as inadequate to have much effect
on incipient inflationary pressures. The

market expected a more forceful move from
the Fed and in the absence of such a
definitive move, revised inflation
expectations upward. Either explanation is
consistent with the increase in market
interest rates that accompanied the Fed’s
tight-ening move.
A third explanation—which does not
involve any revision to expectations of
inflation—seems more plausible, however.
Because the FOMC tends to move the federal funds rate in a series of increments,
the increase on February 4, 1994 led
market participants to anticipate further
increases. As a result, long-term rates,
which reflect current and expected shortterm rates, increased by more than the
federal funds rate.
Figure 5 supports the notion that the
25 basis point increase on February 4 led
market participants to expect further
increases. At the same time that the expected funds rate for February rose in
anticipation of the move on February 4,
the implied three-month future yield also
rose. By the time the February increase in
the federal funds rate was announced, the
futures market was already predicting
another 25 basis point increase within the
next three months. This expectation was
mirrored in the comments of market analysts at the time: for example, one market
observer interpreted the funds rate increase as “the first step on a journey that is
going to last some time.”9
So the behavior of market rates at the
time of the Fed’s first move to tighten
policy could have been caused by an awakening of inflation fears, by the arbitrage
effect of current and prospective increases
in the federal funds rate, or conceivably by
some combination of these effects.

7

Statement before the Joint
Economic Committee, United
States Congress, January 31,
1994. Federal Reserve Bulletin
(March 1994, p. 233).

8

Joseph Liro, chief economist at
S.G. Warburg, quoted by
Thomas D. Laurencella and
Laura Young, Wall Street
Journal, February 1, 1994,
p. C23.

9

John Lipsky, chief economist at
Salomon Brothers, quoted by
Thomas T. Vogel, Wall Street
Journal, February 7, 1994,
p. C19.

May 1994

After two more increases of 25 basis
points each in March and April, the FOMC
raised its objective for the federal funds
rate by 50 basis points on May 17, 1994.
The response in the bond market was the
reverse of previous funds rate increases. As
the May FOMC meeting approached, longterm bond yields declined. After the funds

FEDERAL RESERVE BANK

27

OF

S T. L O U I S

J U LY /A U G U S T 1996

futures market was predicting a high prob(January
1996)
ability
of 1959-June
a 50-basis-point
increase.
Did the magnitude of the funds rate
increase convince market participants that
the Fed’s anti-inflation strategy would be
successful? That is one explanation of the
decline in bond yields. That conclusion,
however, cannot be drawn with certainty.
Once again, the expectations hypothesis
suggests an alternate, though not mutually
exclusive, interpretation. After the 50basis-point increase, there was speculation
that the FOMC would not have cause to
raise the funds rate again in the near
future. The official statement released by
the FOMC following its meeting contributed to this sentiment: “These actions . . .
substantially remove the degree of monetary accommodation which prevailed
throughout 1993.”11 A Wall Street Journal
writer
interpreted
this statement as being
(January
1959-June 1996)
quite clear: “Yesterday’s declaration means
that the Fed now believes it is very close to
neutral and doesn’t expect any further rate
increases soon.”12 To the extent that bond
market participants lowered their expectations of further increases in the funds rate,
the expectations theory of interest rates
would predict a decline in bond yields,
even if inflation expectations remained
unchanged.
The reaction of the federal funds
futures markets gives some credence to
this view. As shown in Figure 5, the
implied rate on three-month futures was
falling for a period both before and after
the meeting. Nevertheless, it continued to
indicate that at least one more increase of
25 basis points was likely within the next
three months. Hence it is unclear whether
the bond market’s reaction to the policy
move on May 17, 1994, indicated a reduction in expected inflation, a change in
the short-term outlook for Fed policy,
or both.

Figure 6

The Market Response to Changes in the
Fed Funds Target
7.0

Interest Rates

7.0
-25

6.5

6.5

6.0

6.0

5.5

5.5

5.0

5.0

4.5

4.5

4.0
4.0
1/2/96 1/11/96 1/23/96 2/1/96 2/12/96 2/22/96 3/4/96 3/13/96 3/22/96
31 Jan 96
Ten Year T-Note

Three Year T-Note

One Year T-Bill

Three Month T-Bill

Figure 7

Fed Funds Futures Market Implied
Expected Funds Rate
5.75

Percent

Percent

Target before 1/31/96

5.50

5.50

Expected Rate
Target after 1/31/96

5.25

5.25

Expected Rate

Expected Rate
5.00

5.00

4.75

4.75
Jan

10

5.75

Dave Kansas, Wall Street
Journal, May 17, 1994, p. C2.

11

Federal Reserve Bulletin, July
1994, p. 610.

12

David Wessel, Wall Street
Journal, May 18, 1994, p. A3.

Feb

Mar

Apr

May

June

July

rate increase was announced, bond yields
continued to decline. On the day of the
funds rate change, the yield on 10-year
Treasury notes, for example, fell by 21
basis points.
The decline in bond yields appears to
have been directly related to the Fed’s
move. Reports in the financial press
suggest that there was a great deal of
uncertainty about the timing and
magnitude of the policy move. On the
morning of the meeting, a Wall Street
Journal reporter noted that “several
interest-rate watchers expect an increase in
rates. The only question is how much?”10
Figure 5 shows that the federal funds

January 1996

A third example serves to show the
dynamic nature of market expectations
and their responses to Federal Reserve
policy. On January 31, 1996, the Fed
voted, in effect, to reduce its target for the

FEDERAL RESERVE BANK

28

OF

S T. L O U I S

J U LY /A U G U S T 1996

federal funds rate by 25 basis points, from
5.50 percent to 5.25 percent. (At the same
time, the Fed lowered the discount rate
from 5.25 percent to 5.00 percent.) According to the financial press, the Fed’s
action was widely expected and the rise in
short-term security prices in preceding
days reflected anticipation of the move.13
Between January 1 and January 30, 1996,
market yields on short-term Treasury securities fell some 20 to 30 basis points. The
yields on government securities with
maturities of seven years or more, however, did not fall over the period.
Government security yields did
decline, but only modestly, after the Fed’s
cut in its funds rate target on January 31.
Although the Fed reduced its target by 25
basis points, market yield declines ranged
from eight basis points on three-month
bills to just one basis point on 30-year
bonds. Yields on short- and medium-term
securities continued to decline through
mid-February, however, but those on longterm government securities changed little—
some even increased. Then, from midFebruary through March, yields on all
securities rose. For illustration, the daily
yields on three-month, one-year, threeyear and 10-year Treasury securities are
plotted in Figure 6.
How might we interpret the behavior
of interest rates both before and after the
Fed’s reduction in its funds rate target on
January 31, 1996?
The modest changes in interest rates
that occurred on January 31, support the
press’s view that the Fed’s action had been
widely anticipated. Further evidence of
this can be seen in Figure 7, which plots
the expected average federal funds rate in
different months using data from the federal funds futures market. On January 30,
the funds rate the market expected to prevail during February lay between the prevailing Fed target of 5.50 percent and the
new target of 5.25 percent established on
January 31. That the expected rate lay
closer to the new target indicates that on
January 30 the market believed that the
Fed was more likely than not to reduce
its target on January 31. When the Fed

validated these expectations, the expected
funds rate for February fell immediately to
5.25 percent.
The data charted in Figure 7 also illustrate that on January 30 the futures market
expected not only the funds rate cut on
January 31, but also further cuts from
March through July. After the Fed reduced its target, these expectations only
hardened.
Further evidence that the Fed’s action
on January 31 was widely anticipated is
reflected in the lack of change in intermediate- and long-term Treasury security
yields on that date. The failure of longterm yields to change significantly on the
Fed’s easing move is thus consistent with
the behavior of short-term rates, the
federal funds futures market, and the
financial press, all of which suggest
that the Fed’s move was widely anticipated.
Between mid-February and March 31,
1996, market interest rates generally rose.
As illustrated in Figure 6, rates made two
especially large jumps in mid- and late
February and one more in early March.
Throughout the period, new data suggested that the economy was growing
more quickly than some previously released indicators had suggested. Moreover,
in mid-February, rising commodity prices
suggested to some market participants that
inflation was likely to rise, causing market
security yields to rise.14 Although yields
rose across the spectrum of maturities,
long-term security yields rose most. This
pattern of rate changes suggests that the
new information caused market participants to revise their expectations of the
Fed’s target for the federal funds rate upward over ensuing months, and possibly
expectations of inflation as well.
Market interest rates again rose when
Federal Reserve Chairman Greenspan
testified before Congress about monetary
policy and the state of the economy on
February 20, 1996, which many analysts
interpreted as confirmation that additional
funds rate reductions over the near term
were unlikely. Finally, the release of new
employment data on March 8, 1996,

FEDERAL RESERVE BANK

29

OF

S T. L O U I S

13

See, for example, Dave
Kansas, Wall Street Journal,
January 31, 1996, p. C1.

14

For example, see Dave Kansas,
Wall Street Journal, February
15, 1996, p. C1.

J U LY /A U G U S T 1996

revealing an unexpectedly large increase in
employment during February is widely
cited for a sharp increase in bond yields on
that date. According to one report, “The
carnage [in the bond market] began immediately after a stronger-than-expected
employment report snuffed out hope that
Federal Reserve policymakers would lower
short-term interest rates anytime soon.”15
The evolution of expectations about
the course of Fed policy was reflected in the
federal funds futures market. In addition to
the expected future funds rate path implied
by market pricing on January 30 and
January 31, Figure 7 plots the implied path
based on futures market data from March 8.
In contrast to the earlier dates, when further
funds rate cuts were expected, on March 8
the market expected the funds rate to
remain at 5.25 percent through July 1996.
According to the expectations hypothesis, the rise in long-term interest rates on
March 8 reflected the expectation that
short-term rates would rise in the future.
The increase in long-term rates could also
reflect a revised anticipation of higher
inflation in the future, though other explanations, such as an increase in the real
interest rate, could also explain the rise.
Inevitably, because many factors affect the
supply of and demand for securities, any
one move in market yields can have
several non–mutually exclusive explanations. Nevertheless, the behavior of market
rates after January 31, 1996, is consistent
with, first, a period of relative calm in
which markets anticipated further reductions in the Fed’s interest rate target, with
little apparent change in inflation expectations. Then, following new information
about the health of the economy and new
speculation about Fed behavior, markets
changed their expectations about the nearterm course of monetary policy and perhaps revised their expectations of future
inflation upward.

15

Vogelstein and Jereski, Wall
Street Journal, March 11,
1996, p. C1.

can be difficult. Sometimes market rates
rise when the Fed’s target is raised, and
sometimes they fall. Sometimes rates move
by more than the change in the funds rate
and sometimes by less. These responses
can be interpreted as an amalgam of inflation expectations, anticipated future
monetary policy actions, and changes in
real rates of return.
Although these influences are difficult
to disentangle, the information from the
federal funds futures market can help
identify the role of expectations in the
determination of market interest rates.
Specifically, with an understanding of the
extent to which a Fed policy action is anticipated in financial markets, we can better
interpret subsequent changes in market
interest rates.
Throughout 1994 and 1995, however,
the behavior of the federal funds futures
market suggests that most Fed actions
were at least partly anticipated. Moreover,
the Fed’s tendency to move its target for
the federal funds rate incrementally in one
direction before reversing course is built
into market expectations of future policy
actions, as revealed in both the spot
markets for Treasury securities and the
federal funds futures market. The incremental nature of Fed policy moves, along
with interest rate arbitrage, likely also
explains why market interest rates typically moved in the same direction as
changes in the federal funds rate during
1994–95. When a policy move is widely
anticipated, and particularly if it is expected to be one of many in a series of
moves in the same direction, market expectations about inflation are not altered. Only
surprise moves, or moves that are widely
taken as turning points, will typically alter
expectations about inflation.

REFERENCES
Campbell, John Y. “Some Lessons From the Yield Curve,” Journal of
Economic Perspectives (Summer 1995), pp. 129–52.
Carlson, John B., Jean M. McIntire and James B. Thomson. “Federal
Funds Futures as an Indicator of Future Monetary Policy: A Primer,”
Federal Reserve Bank of Cleveland Economic Review, vol. 31, no. 1
(1995 Quarter 1), pp. 20–30.

CONCLUSION
Evaluating the credibility of monetary
policy by observing bond market reactions

FEDERAL RESERVE BANK

30

OF

S T. L O U I S

J U LY /A U G U S T 1996

Carlstrom, Charles T. “A Monetary Policy Paradox,” Economic
Commentary, Federal Reserve Bank of Cleveland (August 15, 1995).
Chicago Board of Trade, “Flexible Futures For Managing Interest Rate
Risk,” 1995.
Dotsey, Michael, and Jed L. Devaro. “Was the Disinflation of the Early
1980s Anticipated?” Federal Reserve Bank of Richmond Quarterly
Review (Fall 1995), pp. 41–59.
Dueker, Michael. “When Are Low-Inflation Policies Credible?” Federal
Reserve Bank of St. Louis Monetary Trends, January 1996.
Estrella, Arturo, and Frederic S. Mishkin. “The Term Structure of Interest
Rates and Its Role in Monetary Policy for the European Central Bank”
NBER Working Paper #5279, 1995.
Fama, Eugene F. “Term-Structure Forecasts of Interest Rates, Inflation,
and Real Returns,” Journal of Monetary Economics (January 1990),
pp. 59–76.

Federal Reserve Bulletin, various issues.
Gavin, William T. “The FOMC in 1995: A Step Closer to Inflation
Targeting?” this Review (forthcoming).
Krueger, Joel T., and Kenneth N. Kuttner. “The Fed Funds Futures Rate
as a Predictor of Federal Reserve Policy,” Federal Reserve Bank of
Chicago, working paper WP-95-4, March 1995.
Mishkin, Frederic S. “What Does the Term Structure Tell Us About Future
Inflation?” Journal of Monetary Economics (January 1990),
pp. 77–95.
Pakko, Michael R. “The FOMC in 1993 and 1994: Monetary Policy in
Transition,” this Review (March/April 1995), pp. 3–25.
Rudebusch, Glenn D. “Do Measures of Monetary Policy in a VAR Make
Sense?” working paper, Federal Reserve Bank of San Francisco
(March 1996), #96-05.

The Wall Street Journal, various issues.
Thornton, Daniel L. “Does the Fed’s New Policy of Immediate Disclosure
Affect the Market?” this Review (forthcoming).

FEDERAL RESERVE BANK

31

OF

S T. L O U I S

J U LY /A U G U S T 1996

Appendix

CALCULATIONS
UNDERLYING FIGURE 5

date and the FOMC meeting. This is the
calculation underlying the expected funds
rate measures illustrated in Figure 5 for
days leading up to FOMC meetings. For
the days following the meeting, the
following for days leading up to FOMC
meetings. For the days following the
meeting, the following more simple
formula

Figure 5 presents estimates of expected FOMC policy actions, as derived
from the federal funds futures market. To
isolate the funds rate that is expected to
prevail after an FOMC meeting, some calculations are necessary. At any point
during a month, the current-month federal
funds futures rate (i f ) can be thought of as
a weighted average of two components—
the actual funds rate experienced to date
(ia), and the rate expected to prevail for
the rest of the month (im):
(1)

if =

(4)

ie =

is used.‡

T
N−T
× ia +
× im ,
N
N

where T is the number of days passed to
date and N is the number of days in the
month. This equation can be solved for the
rest-of-month expected rate.
If there is a meeting of the FOMC,
however, then the expected rate for the
rest of the month can be similarly expressed as a weighted average of two
components—the prevailing federal funds
target, (i*) and the rate expected to prevail
after the meeting:
(2)

im =

M−T
N−M
× i* +
× ie ,
N−T
N−T

where M is the FOMC meeting date.
Combining these two expressions and
solving for ie gives the following:
‡

To prevent distortions that
sometimes appear toward the
end of the month (because of
the nature of the futures contract), the implied funds rate
from the subsequent month is
used for days following FOMC
meetings in cases where the
meeting date falls within the
last five business days of a
month. See the shaded box,
p. 24.

(3) ie =

(

)

N × i f − T × ia − M − T × i*
N−M

.

Hence we can find the rate expected to
prevail following an FOMC meeting by
taking the rate implied in the current
futures contract and subtracting
components related to the actual funds
rate to date and the target funds rate
expected to prevail between the current
FEDERAL RESERVE BANK

32

OF

S T. L O U I S

N × i f − T × ia
N−T

,

J U LY /A U G U S T 1996

Christopher J. Neely is a research economist at the Federal Reserve Bank of St. Louis. Kent A. Koch provided research assistance.

The Giant
Sucking Sound:
Did NAFTA
Devour the
Mexican Peso?

This article examines the relationship
between NAFTA and the peso crisis of
December 1994. First, the provisions of
NAFTA are reviewed, and then the links
between NAFTA and the peso crisis are
examined. Despite a blizzard of innuendo
and intimation that there was an obvious
link between the passage of NAFTA and
the peso devaluation, NAFTA’s critics
have not been clear as to what the link
actually was. Examination of their arguments and economic theory suggests two
possibilities: that NAFTA caused the Mexican authorities to manipulate and prop
up the value of the peso for political reasons or that NAFTA’s implementation
caused capital flows that brought the
peso down. Each hypothesis is investigated in turn.

Christopher J. Neely

A

t the end of 1993 Mexico was touted
as a model for developing countries.
Five years of prudent fiscal and monetary policy had dramatically lowered its
budget deficit and inflation rate and the
government had privatized many enterprises that were formerly state-owned. To
culminate this progress, Mexico was
preparing to enter into the North American
Free Trade Agreement (NAFTA) with
Canada and the United States. But less than
a year later, in December 1994, investors
sold their peso assets, the value of the Mexican peso plunged 50 percent against the
U.S. Dollar, and Mexico was forced to borrow from the International Monetary Fund
(IMF) and the United States to get through
a financial crisis. In 1995, inflation in Mexico soared to 50 percent and real gross domestic product (GDP) fell by 4 percent.
Politicians and commentators like Ross
Perot, Pat Buchanan, William Greider, and
Robert Kuttner blamed the enactment of
NAFTA for the devaluation of the peso and
the ensuing economic turmoil in Mexico,
with some calling for its renegotiation or
even repeal. As the members of NAFTA
consider expanding to encompass other
Latin American nations, such as Chile, investors and policymakers should understand the link between NAFTA and the
peso crisis well. Did NAFTA cause or exacerbate the devaluation of the peso? Or did
NAFTA help alleviate some of the consequences of the crisis?

NAFTA
NAFTA grew out of the U.S.–Canadian
Free Trade Agreement of 1988.1 It was
signed by Mexico, Canada, and the United
States on December 17, 1992. The legislatures of those countries ratified NAFTA,
and the agreement took effect on January
1, 1994. The treaty substantially lowered
national barriers to trade and investment
in North America, giving consumers more
choices and lower prices. In addition, the
changes began to lower the cost of production and to funnel investment and labor to
their most productive uses. Not surprisingly, the costs—real and imagined—of
this reallocation of resources stirred the
passions of those opposing the agreement.
The trade provisions of NAFTA were
designed to reduce tariffs and nontariff
barriers—such as quotas and import
licensing—radically over 15 years. Some
tariffs were reduced immediately, whereas
other reductions will be phased in over a
period of 10 years—15 years for certain
sensitive sectors, such as agriculture and
textiles and apparel.

FEDERAL RESERVE BANK

33

OF

S T. L O U I S

1

See Hufbauer and Schott
(1993), Aguilar (1993), or
Tornell and Esquivel (1995)
for more discussion of NAFTA’s
provisions.

J U LY /A U G U S T 1996

2

See Tornell and Esquivel
(1995). Changes in valueweighted tariff schedules can
be misleading, however, because there are also some
quantitative restrictions.

3

See Kehoe (1995) for a review of Mexico’s recent trade
history.

4

The GATT was an international
organization to negotiate free
trade among its members. It
has been superseded by the
World Trade Organization
(WTO).

5

Krugman (1993) and Brown,
Deardorf and Stern (1992) discuss estimates of the gains
from NAFTA.

Despite the impressive achievements
of the negotiators in crafting such a farreaching trade agreement, NAFTA’s direct
economic benefit to the United States will
likely be small. One representative estimate of NAFTA’s annual benefits to Mexico and the United States arrives at approximately the same figure for each
country;5 however, this amounts to about
0.3 percent of 1993 U.S. GDP but more
than 5.0 percent of Mexico’s output. Schott
(1994), Tornell and Esquivel (1995) and
others have argued that the most important aspect of NAFTA’s passage for the
Mexican economy is that it would cement
the other economic reforms in place. Krugman (1993) and Orme (1993) both contend
that NAFTA is most important to the United
States as a tool of foreign policy, to encourage Mexican economic and political reform.

For the United States and Mexico, the
trade provisions of NAFTA are expected to
have their most important effects on the
automobile, textile and apparel, and agricultural sectors. In agriculture, U.S. and
Mexican quotas were immediately converted into equivalent tariffs and those
tariffs will be phased out over 10 to 15
years. As Hufbauer and Schott (1993)
note, this is a remarkable achievement
given the difficulties encountered by
other free trade agreements on agricultural issues.
Given the fierce fight in the United
States over the agreement, it is ironic that
NAFTA required more substantial changes
in Mexican law—both trade and investment law—than it did in U.S. law. Average
U.S. tariff levels on Mexican goods were
already quite low—just four percent—on a
value-weighted basis, before the introduction of NAFTA.2 Mexican tariffs were
higher, averaging 10 percent on imports
from the United States. Through NAFTA,
Mexico also committed itself to address
other long-standing U.S. concerns, like the
protection of intellectual property rights
and reform of Mexico’s regulation of foreign investment.
NAFTA was the culmination of a significant break with Mexico’s protectionist
past.3 Until the 1970s, Mexico followed a
policy of import substitution industrialization that mandated highly protected markets for manufactured goods. In that
decade, preliminary reforms in the direction of freer trade were taken. The debt
crisis of 1982 reversed that trend; for a
short period in 1982–1983, Mexico was
one of the most protected economies in
the world. During the de la Madrid administration (1982–88), Mexico took important steps to move toward more liberal
trade. Mexico lowered tariffs and joined
the General Agreement on Tariffs and
Trade (GATT) in 1986.4 Mexico took further unilateral steps toward free trade as
part of the Salinas administration’s
(1988–94) program of economic reform.
This period is known as la apertura (the
opening).

NAFTA AND THE VALUE
OF THE PESO
This section lays out the case that the
peso was kept overvalued because of the politics of NAFTA and then investigates whether
this argument is consistent with the facts.

The Case That the Peso’s Value
Was Artificially Inflated Because
of the Politics of NAFTA
The most common hypothesis linking
NAFTA to the peso crisis is that the politics
of NAFTA caused the Bank of Mexico to
systematically manipulate the value of the
peso to increase support for the treaty, both
before NAFTA was passed in the United
States and during its first year. There are
two versions of this hypothesis. The first
version suggests that the value of the peso
was deliberately manipulated to secure political support for NAFTA and that the devaluation—to obtain a trade advantage—
was planned well in advance. The second
version is less sinister. It suggests only that
the Mexican authorities were sensitive to
U.S. politics in setting exchange rate policy
after NAFTA was passed. The following sections lay out the arguments behind each
version of this hypothesis.

FEDERAL RESERVE BANK

34

OF

S T. L O U I S

J U LY /A U G U S T 1996

Deliberate Manipulation and Planned
Devaluation.

of the evidence at that time suggested that
the treaty had cost American jobs. So there
was considerable pressure to produce evidence that showed that NAFTA would create jobs in the United States.
The Mexican government was not immune to such pressure. In 1993, passage
of NAFTA by the U.S. Congress was the
main policy concern of the Mexican administration [see Tornell and Esquivel
(1995)]. In August of that year, President
Salinas even promised to raise the Mexican
minimum wage to alleviate U.S. fears of
cheap Mexican labor driving down U.S.
wages and taking jobs. Critics charge that
because of such political considerations,
the Mexican government deliberately kept
the peso overvalued throughout 1993 and
1994 and planned the eventual devaluation well in advance.

“... the devaluation of the peso
had been planned for more than a
year and was openly discussed at
the highest levels of the Mexican
government. It was also widely
known in Washington. I discussed
it in my testimony before the
House Committee on Small Business in March, 1993—eight
months before the North American Free Trade Agreement was
passed into law.”
Ross Perot, Los Angeles Times,
January 4, 1995.6
Critics like Ross Perot argue that the
Mexican government and the Bank of Mexico kept the value of the peso artificially
high to increase political support for the
treaty in the United States by creating a bilateral trade surplus with Mexico. The
United States did have a trade surplus with
Mexico in the early 1990s. A study by
Hufbauer and Schott (1993) was frequently
cited by NAFTA proponents to support the
questionable notion that the growth of this
trade surplus would create 170,000 jobs in
the United States. The Clinton administration used these arguments to sell NAFTA to
the U.S. Congress primarily as a jobs program, rather than as a trade agreement that
would promote greater choice and lower
prices for consumers and greater efficiency
in production.

Sensitivity to U.S. Politics. A more reasonable hypothesis is put forward by Velasco (1995) and others. They suggest
only that, after NAFTA was passed, the
Mexican authorities were sensitive to the
U.S. political situation and may therefore
have been more reluctant to permit the
peso to depreciate than they would otherwise have been. Specifically, in March
1994, the peso came under speculative
pressure in the wake of the assassination
of Luis Donaldo Colosio, presidential candidate of the ruling Revolutionary Institutional Party (PRI). At that time, a number
of observers warned that the peso was
overvalued and that a faster devaluation
was warranted. Velasco suggests that because such a course of action threatened to
create political problems with the United
States, political exigencies may have prevented an earlier, milder correction to the
value of the peso that would have avoided
the drastic correction of the later crisis.

“We will make our case as hard
and as well as we can. And, though
the fight will be difficult, I deeply
believe we will win. And I’d like
to tell you why. First of all, because
NAFTA means jobs. American jobs,
and good-paying American jobs. If I
didn’t believe that, I wouldn’t support this agreement.”
President Bill Clinton at the signing
of NAFTA Side Agreements on
September 14, 1993.

Evaluating the Case that the Peso’s
Value was Artificially Inflated
Because of the Politics of NAFTA
Critics argue that NAFTA provided the
impetus for the Mexican monetary authorities to maintain the value of the peso in

President Clinton even talked about
leaving NAFTA after three years if a review

FEDERAL RESERVE BANK

35

OF

S T. L O U I S

6

See also, columnist Robert
Kuttner, January 22, 1995,
in the Akron Beacon Journal
and author William Greider in
Rolling Stone, March 9, 1995.

J U LY /A U G U S T 1996

excess of its equilibrium value. The authorities allegedly knew that the peso was
overvalued but gambled that this overvaluation could be maintained long enough to
secure NAFTA’s passage in the United
States. Thus, this hypothesis requires that:

barrel of oil costs $20 in the United States
and 80 pesos in Mexico, the law of one
price predicts the nominal exchange rate
will be $0.25 per peso. This condition
must approximately hold, or people could
make money by buying oil in the country
where it is cheap and selling it in the
country where it is expensive. Such arbitrage would tend to drive the price of oil
down in the country where it is expensive
and raise the price in the country where it
is cheap, until the law of one price approximately holds.
If the law of one price holds for
each good in a price index and the
weights in the price index are the same
for each country, then consumption baskets should also sell for the same price
when expressed in the same currency.
This is called absolute purchasing power
parity (PPP), which can be expressed as
follows:

1. The peso was overvalued.
2. The Mexican authorities knew that it
was overvalued.
3. The Mexican authorities kept it overvalued to increase or at least maintain support for NAFTA in the United States.
Although it is not possible to test the elements of this hypothesis, they may be examined to see whether they are consistent
with the facts. This section argues that
though the evidence favors the view that
the peso was overvalued, this was not obvious at the time. Further, to the extent that
the peso may have been overvalued, the
overvaluation was a result of the disinflation strategy of the Mexican authorities,
rather than a result of NAFTA.

MEX
pU.S.
(t) = pIndex
(t) × e(t),
Index

where pU.S.
(t) is a measure of the price
Index
MEX
(t) is the
level in the United States and pIndex
analogous measure for Mexico. Of course,
because of different patterns of consumption across countries, the presence of nontraded goods and differentiated goods, absolute PPP does not describe the relation of
price levels and exchange rates very well.
A less stringent, but more realistic relation is relative PPP. It says that differences in countries’ inflation rates should
be reflected in changes in the exchange
rate, so that

Nominal and Real Exchange Rates. When
discussing the value of the peso, it is important to distinguish between the nominal
exchange rate, or the price of a peso in
terms of dollars, and the real exchange
rate, the price of Mexican goods in terms
of U.S. goods. This section explains the relationship between prices and exchange
rates and why the real exchange rate is the
relevant measure of the proper value of the
peso.
Exchange rates and prices are linked
through the law of one price, which says that
identical goods should sell for the same
price when expressed in terms of the same
currency.7 In the case of oil, for example,

∆pIndex (t) − ∆pIndex (t) = ∆e(t),
U.S.

where ∆ stands for the percentage change in
a variable over time. This equation says that
if inflation in Mexico exceeds inflation in
the United States, the exchange rate will fall
to reflect the difference. That is, the peso
will depreciate. Why? If Mexican goods become more expensive than U.S. goods, consumers in both the United States and Mexico will tend to buy more U.S. goods. This
will cause the peso to depreciate until Mexican goods are competitive again.

pU.S.
(t) = pMEX
(t) × e(t),
oil
oil

7

Barriers to trade, transportation
costs, and imperfectly competitive markets may prevent the
law of one price from holding.

MEX

where the variable pU.S.
(t) is the price of oil
oil
in dollars in the United States at time t,
pMEX
(t) is the price of oil in pesos in Mexoil
ico at time t, and e(t) is the exchange rate
in dollars per peso. In other words, if a

FEDERAL RESERVE BANK

36

OF

S T. L O U I S

J U LY /A U G U S T 1996

A strict interpretation of relative PPP
says that the real exchange rate, or the
price of Mexican goods in terms of U.S.
goods, should be constant over time. At
time t, the real exchange rate (q(t)) can be
expressed as follows:

Figure 1

Index of the Real Exchange Rate (WPI)
Mean = 100
140

pIndex (t) × e(t)
q(t) = }}
.
pU.S.
Index(t)

120

For practical purposes, however, relative
PPP is interpreted to mean that the real
exchange rate should tend to come back to
its historical average rather than be constant over time. Empirical studies suggest
that this interpretation of relative PPP is
useful for thinking about long-run tendencies in exchange rates, especially when inflation rates are high.
Why is the real exchange rate important? A currency has value only because of
what it can purchase. The real exchange
rate adjusts the nominal value of a currency for its purchasing power and so determines competitiveness in world markets. For example, a rise in the real
exchange rate (as defined previously)
means that the price of Mexican goods in
terms of U.S. goods has risen. The price of
Mexican exports to the United States rises,
hurting Mexican exporters, but imports
from the United States become cheaper to
Mexican consumers. Therefore the relevant measure of the value of the peso is
the value of the real exchange rate.

80

MEX

100

60
40

1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995

Year

of the real exchange rate. If the Mexican
inflation rate minus the U.S. inflation rate
exceeds the rate of depreciation of the
peso—that is, if

∆pIndex (t) − ∆pIndex (t) > −∆e(t),
MEX

U.S.

—then the real exchange rate rises and
Mexican goods became more expensive in
terms of U.S. goods; the peso becomes
overvalued. Historical measures of the correct value of the real exchange rate are imperfect, though. The proper value of the
real exchange rate can change over time
because of changes in productivity, preferences, legal capital controls, or other factors. These changes are usually slow, however, leaving historical measures useful.
Respected economists like Dornbusch
and Werner (1994) argued during 1993
and 1994 that the peso was overvalued because an index of the real exchange rate,
as measured by the Wholesale Price Index
(WPI), was high by historical standards.
As illustrated in Figure 1, this index rose
steadily from a level of 70 in 1987 to a
peak of about 130 at the end of 1993. By
this measure, Mexican goods had become
almost twice as expensive in terms of U.S.
goods from 1987 through 1993 and the
real value of the peso was 30 percent
higher than its historical average from
1975 through 1993. Dornbusch and
Werner cautioned early in 1994 that

Was the Peso Overvalued? Armed with the
concept of the real exchange rate, it is still
difficult to determine whether the peso was
correctly valued because the real exchange
rate changes over time. In the case of a
pegged exchange rate system like Mexico’s,
a real exchange rate is functionally overvalued or undervalued if the nominal exchange
rate is likely to be forced to change quickly.
That is, the real exchange rate should be
compatible with the commitment to the
pegged nominal exchange rate.
Relative PPP suggests a practical measure of whether the current real exchange
rate is likely to be consistent with the peg:
whether it is in line with historical values

FEDERAL RESERVE BANK

37

OF

S T. L O U I S

J U LY /A U G U S T 1996

Carstens (1995) argue that unit labor costs
are a better way to compute real exchange
rates because they more closely reflect the
relative cost of production in Mexico and
abroad. Further, the real value of the peso
for Mexico’s trade depended not only on
its value vs. the dollar, but also on its
value vs. Mexico’s other trading partners,
and therefore they suggest that multilateral
measure of the real exchange rate is more
appropriate. Figure 2 shows that, by the
beginning of 1994, the multilateral effective real exchange rate index, as measured
by unit labor costs, had also risen substantially—about 60 percent—since 1987 but
was still as low as it had ever been before
1986.8 In fact, it was still slightly below its
historical average for the period 1975–94.
Thus Gil-Diaz and Carstens argued that
this historical measure did not show the
real exchange rate to be overvalued.
Second, because the proper value of
the real exchange rate can change over
time because of productivity changes
and other factors, the Mexican authorities
disagreed with Dornbusch and Werner
about the relevance of historical measures. They asserted that NAFTA and
other economic reforms had raised productivity and had increased the correct
(equilibrium) value of the real exchange
rate; that is, the equilibrium price of
Mexican goods had risen.

Figure 2

Index of Unit Labor Costs
Multilateral Real Exchange Rate
Mean = 100
240
200
160
120
80
40

1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995

Year
Figure 3

International Reserves
Billions of U. S. Dollars
35
30
25
20
15
10
5
0

1988

1989

1990

1991

1992

1993

1994

1995

Year

8

Data for the multilateral real
exchange rate were taken from
Gil-Diaz and Carstens (1995).

“The Mexicans were justifiably
proud of the progress they had
made in bringing down inflation,
by means of the exchange rate
link to the dollar, and did not
want to lose it. I suspect they
thought they were in a new world,
as a result of the economic liberalization and NAFTA.”
Economist Jeffrey Frankel,
Statement to the U.S. Senate Committee on Banking, Housing and
Urban Affairs, March 9, 1995.

this situation was untenable and the
peso should be permitted to depreciate
faster.
The hypothesis that the Mexican authorities deliberately manipulated the
value of the peso requires that these authorities knew the peso was overvalued.
Did they know this? In responding to
Dornbusch and Werner, economists at the
Bank of Mexico contended that the real
exchange rate was not overvalued for several reasons. First, another measure of the
real exchange rate—using unit labor costs
instead of price indices—did not show the
peso to be overvalued. Gil-Diaz and

Also, there was very little pressure on
the peso before March 1994, indicating
that the markets did not believe that the
peso was overvalued. In fact, the Bank of

FEDERAL RESERVE BANK

38

OF

S T. L O U I S

J U LY /A U G U S T 1996

Table 1

Mexican Consumption, Savings, Output, and Inflation

Total Consumption*
Private Consumption
Public Consumption
Total Saving*
Private Saving
Public Saving
Real GDP Growth
Inflation (CPI)

1987

1988

1989

1990

1991

1992

1993

1994

89.3
78.8
10.5
10.7
NA
NA

91.4
81.3
10.1
8.6
7.8
0.8

90.5
80.8
9.7
9.5
7.5
2.0

89.2
79.8
9.5
10.8
6.6
4.2

90.7
80.6
10.1
9.3
5.1
4.2

92.8
81.4
11.4
7.2
3.8
3.5

93.5
81.2
12.3
6.5
NA
NA

95.9
82.3
13.6
4.1
NA
NA

0.0
159.2

1.3
51.7

3.3
19.7

20.9
29.9

9.3
18.8

2.8
11.9

0.4
8.0

3.8
7.1

* Table entries are expressed as a percentage of National Disposable Income.
Source: OCED National Accounts and DRIINTL.

Mexico had to intervene in the market to
sell pesos/buy dollars to keep the value of
the peso down in January 1994, accumulating foreign exchange reserves. Figure 3
shows this accumulation as the spike upwards in foreign exchange reserves at the
beginning of 1994.
Finally, a fundamental measure of
whether the real exchange rate is properly
valued is its effect on exports. The Mexican government questioned how the real
exchange rate could be overvalued when
export growth was as strong as it was. Cumulative nonoil export growth from 1985
to 1994 was more than 200 percent, in
the same range as such export powers as
Hong Kong, Korea, Singapore, and Taiwan.9
To summarize: Dornbusch and Werner
presented evidence that the real exchange
rate, as measured by the WPI, was overvalued in 1993 and 1994. Although in retrospect it looks as if Dornbusch and Werner
were correct, this was not obvious at the
time. Other measures of the exchange rate
showed no overvaluation, economic reform had likely made historical measures
less reliable than usual, and export growth
was strong.

could lead to a crisis.10 These economists argued that the peso had become overvalued
because Mexican officials had used the
pegged exchange rate to help bring inflation
down (see Table 1) from 159 percent in
1987 to 8.0 percent in 1993. This section explains the role of a pegged exchange rate in
bringing down inflation and the dangers of
such a policy.
To understand how the value of the
peso affects inflation, consider how monetary policy, exchange rates, and prices interact. Because only the Bank of Mexico,
Mexico’s central bank, can issue peso currency or reserves, within very broad limits,
it can control the value of the peso by controlling the supply of pesos. Similarly, the
Bank of Mexico also controls Mexican inflation by increasing or decreasing the
growth of the money supply. No central
bank, however, can independently control
both the exchange rate and inflation at
the same time. The desired inflation rate
may not be compatible with the preferred
exchange rate. That is, if a central bank
picks a level of inflation to target, it must
choose the particular path for the exchange rate that is consistent with that
inflation rate. By choosing a path for the
exchange rate (and money growth) consistent with a low inflation rate, the Bank of
Mexico could use a pegged exchange rate
as a tool to help lower the inflation rate.

Disinflation and the Overvalued Peso. In
1993 and 1994 many economists who supported NAFTA warned that the real exchange rate had become overvalued and

FEDERAL RESERVE BANK

39

OF

S T. L O U I S

9

Data taken from Gil-Diaz and
Carstens (1995).

10

See Dornbusch and Werner
(1994) and Hufbauer and
Schott (1993).

J U LY /A U G U S T 1996

11

A crawling peg is a preannounced daily rate of slow
devaluation. In the target zone
system, Mexican authorities
pledged to keep the exchange
rate with the dollar within
given margins.

There are three ways in which a
pegged exchange rate policy helped the
fight against inflation. First, a stronger
peso forced Mexican producers of tradeable
goods to restrain price increases to directly
compete with foreign producers. Second,
in every disinflation, the credibility of the
disinflation is important to breaking the
momentum of the inflation with little cost
in lost output. That is, people have to be
convinced that inflation will fall. A pegged
exchange rate helps break inflationary expectations by providing a concrete measure
of the progress in fighting inflation; it
gives the public an inflation-sensitive nominal anchor. People can see that the currency doesn’t free fall against a (low inflation) foreign currency and so they come to
believe that inflation is falling. Third,
maintaining the exchange rate against the
dollar gives the monetary authority instant
feedback as to the pressure on the value of
the peso.
The danger with using a pegged exchange rate to fight inflation is that the real
exchange rate will become overvalued if
domestic inflation exceeds the rate of depreciation of the domestic currency plus foreign inflation. Pegging the nominal exchange rate while domestic inflation
exceeds foreign inflation raises the real exchange rate, and domestic goods become
more expensive in terms of foreign goods.
This fights domestic inflation for the reasons outlined previously, but at the cost
of making domestic industries less competitive in tradeable goods. Such a situation may quickly become unsustainable.
Despite this danger, many developing
countries with histories of high inflation
have used restrictive monetary policy with
a pegged (or crawling peg) exchange rate
to control inflation. That is the course
Mexico chose; from 1988 to 1994, the
Bank of Mexico used the exchange rate as
an instrument to bring down inflation.
The peso was pegged to the dollar in
March 1988. In January 1989, the peg was
changed to a crawling peg and a moving
target zone was introduced in December
1991.11 The lower limit of the target zone

or band was lowered (devalued) only
slowly. The principle of controlling the exchange rate to restrain inflation remained
the same, however.
As the preceding section concluded, it
was not obvious that the peso was overvalued. To the extent that it may have
been, however, creating an overvalued exchange rate by using a pegged exchange
rate to bring down inflation is neither
new nor unique to Mexico. Numerous authors, including Corbo and De Melo
(1987), have commented on the tendency
toward overvaluation in the so-called
“Southern Cone” countries of Argentina,
Chile and Uruguay when the exchange
rate is used as an instrument to reduce
inflation. Gil-Diaz and Carstens (1995)
add Brazil and Finland to this list of
countries that experienced overvaluation.
In all of these countries, there was substantial real overvaluation but no free
trade agreement to blame for it.
Other Reasons to Avoid Devaluation. Velasco (1995) discusses several reasons why
the Mexican authorities wished to avoid devaluation in 1994. First, they did not wish to
lose the gains they had made against inflation. Aside from the domestic consequences
of loss of control of inflation, the Mexican
authorities feared that a devaluation would
be ineffective in changing the relative price
of Mexican and foreign goods if inflation
would outpace the depreciation of the peso.
Such a devaluation would have been the
worst of both worlds: more inflation, a loss
of credibility and no improvement in the
competitiveness of domestic goods. Further,
to maintain their credibility with investors,
the Mexican policymakers were reluctant to
devalue even in the face of large shocks.
They were concerned that devaluation
would call into question the policymakers’
commitment to other reforms and result in a
loss of foreign investment.
Summary on the Value of the Peso. In
1993–1994 Dornbusch and Werner presented evidence, convincing in retrospect,
that the peso was overvalued. It was not
clear at the time, however, that this was the

FEDERAL RESERVE BANK

40

OF

S T. L O U I S

J U LY /A U G U S T 1996

case. To the extent the peso may have been
overvalued, it was because of the disinflation strategy pursued by Mexico, and other
policy concerns. The evidence is not consistent with the claim that the government of
Mexico deliberately manipulated the value
of the peso and planned a devaluation
years in advance or that the authorities
avoided a faster rate of depreciation solely
(or primarily) because of the politics of
NAFTA.

Figure 4

Current Account and Capital Account
Balances as a Percentage of GDP
15

Capitol
Account

10
5
0
Current
Account

-5

NAFTA AND
INTERNATIONAL
CAPITAL FLOWS

-10
1988

1989

1990

1991

1992

1993

1994

1995

Year

This section introduces the concept of
capital flows, lays out the hypothesis that
NAFTA was responsible for the peso crisis
by stimulating capital flows out of Mexico,
and then shows that the evidence is not
consistent with this hypothesis.

opposite to the capital account balance because a country can import more than it
exports only by selling foreigners claims
on existing real or financial assets.12 Thus
a deficit in the current account must be
balanced by an equal and opposite capital
account surplus because the two accounts
are the opposite sides of the same transaction. One measures the net value of the
goods and services received, and the other
measures the net value of the assets exchanged for the goods and services. A nation that runs a current account deficit
(and, by definition, a capital account surplus) is borrowing from abroad, selling assets like bonds in exchange for new goods
and services. A country running a current
account surplus is lending to other countries by buying assets in exchange for exports of goods and services. In a world
with balanced trade, there would be no
opportunities for net international borrowing, and domestic savings would have to
equal domestic investment.
Figure 4 illustrates that Mexico ran increasing current account deficits and capital account surpluses for the period
1990–1994. In other words, it was increasingly borrowing from abroad—as much as
8 percent of its GDP by 1993. Capital inflows—a capital account surplus—are useful because they permit a nation to consume more and grow faster by borrowing

What Are Capital Flows?
Capital flows entail the buying and
selling of existing assets. When foreign investors buy real or financial Mexican assets, for example, capital flows into Mexico. Real assets include factories and real
estate; financial assets encompass bonds
and equity. Foreign investment is divided
into foreign direct investment (FDI) and
portfolio investment. FDI is distinguished
from portfolio investment by active control of the assets: Buying a factory is FDI,
buying a bond is portfolio investment.
The national income accounts measure
net capital flows by the balance in the capital account. A surplus in a nation’s capital
account means that more capital is flowing
into the country than is flowing out; that
is, the country is selling more existing assets than it is buying. Similarly, the current
account measures trade in goods and services, net receipts on foreign investment,
and unilateral transfers. A current account
deficit means that a country is importing
more newly produced goods and services
than it exports.
Aside from measurement errors, the
current account balance must be equal and

FEDERAL RESERVE BANK

41

OF

S T. L O U I S

12

The accumulation or loss of official reserves like foreign exchange, gold, or other assets
permits an exception to the
rule that the current and capital
accounts must balance. A nation can temporarily finance a
current account deficit by selling off official assets, as Mexico did in 1994. This simply
amounts to a change in the
way that the capital account is
defined.

J U LY /A U G U S T 1996

2. NAFTA triggered capital outflows that
led to the peso’s devaluation by creating political instability.

against future income. The sustainability
of capital inflows (borrowing) is limited by
the capacity of the borrower to pay back
the loan. Borrowing for present consumption is not sustainable unless national income, or the capacity to pay back the
loan, grows rapidly. Borrowing to invest
in productive capacity, borrowing that
increases future income or reduces future
expenditures is more likely to be sustainable. Judging whether capital flows are
sustainable is difficult, however, because
consumption and investment are defined
and measured imperfectly. For example,
spending on education, health care, or
consumer durables is counted in the national accounts as consumption, but perhaps it should be called investment.

The Case that NAFTA Generated Capital
Inflows. The capital inflows to Mexico
(Mexico’s capital account surplus) in
1990–1994 meant that Mexico was borrowing from abroad to finance its current
account deficit. A low savings rate made
Mexico more dependent on international
capital flows and therefore more vulnerable to shocks.13 Critics contend that this
dependence was critically worsened by
passage of NAFTA. There are two ways in
which NAFTA might generate capital inflows to Mexico. The first is by decreasing
Mexican national savings. The second is
by increasing the desirability of investment
in Mexico.
Why might NAFTA reduce Mexico’s
savings rate? First, by directly lowering
trade barriers, NAFTA made consumption
of imports, especially consumer durables,
cheaper and more attractive relative to saving. Given Mexico’s history of protectionism, consumers may have feared that free
trade was temporary and wished to buy
while they could. A rise in the consumption rate must lower the savings rate because all disposable income of a nation or
an individual can be classified as either
consumption or savings. Second, NAFTA
and other economic reforms may have increased expectations of future income, increasing Mexicans’ willingness to go into
debt and lenders willingness to permit
this.14 At the same time, financial reforms
gave ordinary people greater access to
credit markets and thus greater ability to
go into debt. Finally, if NAFTA contributed to an artificially higher real value
of the peso, that would have also made
imported goods much less expensive and
consumption more attractive.

The Case that NAFTA Was
Responsible for Capital Flows
that Caused the Peso Crisis

13

A savings rate and a consumption rate are savings and consumption, respectively, as percentages of income.

14

The Permanent Income Hypothesis, developed by economist
Milton Friedman (1957), predicts that people base their
consumption on their lifetime
income. That is, they smooth
their consumption over time by
borrowing during periods of
low income and saving during
periods of high income.

The immediate precipitating factor in
the Mexican peso crisis of December 1994
was the desire of investors to get their assets, especially portfolio investment, out
of pesos, which they feared would be
devalued, and into dollars or other foreign
currency. That is, capital was flowing out
of Mexico. This section lays out the logic
behind the critics’ second hypothesis
about NAFTA and the Mexican financial
crisis—that NAFTA drove international
capital flows that led to the devaluation of
the peso. There are also two versions of
this hypothesis. The first version requires
only that NAFTA simply encouraged
capital inflows—either by depressing national savings or by making Mexico a
more attractive investment environment—
and that capital inflows, in the form of
portfolio investment, are inherently dangerous. The second version suggests that
NAFTA generated political instability
that sparked capital outflows and the
devaluation. These hypotheses require
that:

“... NAFTA served as a kind of
‘Good Housekeeping Seal of Approval’ that encouraged even more
investors into Mexico.”
Anderson, Cavanagh and Ranney
(1996), p. 3.

1. Either NAFTA encouraged international
capital inflows, which are intrinsically
destabilizing, or
FEDERAL RESERVE BANK

42

OF

S T. L O U I S

J U LY /A U G U S T 1996

The second form of the capital inflow
hypothesis suggests that NAFTA may have
generated capital inflows to Mexico by making Mexico a more attractive investment environment. This hypothesis would explain
the surge, in early 1994, of capital inflows
that caused the peso to appreciate. NAFTA
was considered especially important to investors because an international treaty
made the reforms more likely to be permanent. There is considerable reluctance
to break a treaty with a foreign government.
An implicit assumption of the hypothesis that NAFTA was responsible for the peso
crisis because it encouraged capital inflows
is that such flows are inherently destabilizing. Portfolio investment, in particular, was
frequently maligned as being a cause of the
crisis. It was said to be moved on a whim
with a short-term investment horizon, creating financial market volatility. Such a view
requires that international capital markets
be subject to fads or speculative bubbles.
Critics point to the volatility of the dollar in
the 1980s, the European Exchange Rate
Mechanism crises of 1992 and 1993, and the
recent flood of capital into emerging markets as evidence of this.

tween Mexico’s prosperous north
and an impoverished south.”
The Economist, January 8, 1994.
The uprising was soon contained by
the Mexican army, but it and other political
shocks concerned investors throughout the
year. They engendered fears that the economic reforms in Mexico had moved too
fast and would lead to social unrest that
would roll back the reforms. In fact, the initial devaluation on December 20, 1994, was
sparked by a run on the peso started by rumors of renewed fighting in Chiapas.15
These political shocks led investors to
exchange pesos for dollars at the Bank of
Mexico, causing a series of falls in Mexico’s
foreign exchange reserves, limiting its
short-term ability to defend the peso.16 Figure 3 illustrates the stepwise falls in foreign
exchange reserves during 1994. By the time
that rumors of renewed fighting rattled the
markets on December 19, 1994, the Bank
of Mexico had nearly run out of foreign exchange reserves. Without foreign exchange
to defend the peso, the Bank of Mexico had
to devalue.17 Critics of NAFTA might argue
that the treaty caused the peso crisis by
sparking the Chiapas uprising.

The Case that NAFTA Contributed to Capital Outflows Through Political Instability.
From the Mexican view, the purpose of
NAFTA was to create a more prosperous
and stable Mexico. Nevertheless, even good
economic policy can unintentionally create
dislocations and political instability. Some
have charged that NAFTA contributed to
the Chiapas uprising that triggered the
capital outflows that brought down the
peso.

15

See Gil-Diaz and Carstens
(1995) or IMF (1995) for the
details of the decision to devalue.

16

In the long run, the Bank of
Mexico used its control over
the money supply to determine
the foreign exchange value of
the peso. Over the short term,
however, the Bank of Mexico
defended the value of the peso
by buying and selling pesos for
dollars. By itself, this action
would reduce the supply of
pesos and push up Mexican interest rates. The Bank of Mexico, however, fully sterilized
the purchase of pesos by buying outstanding bonds in exchange for pesos, putting the
pesos back into circulation.
Sterilization is intended to
leave domestic interest rates
unchanged after foreign exchange purchases or sales.

17

Some suggest that the Bank of
Mexico could have used its
control over the domestic
money supply to defend the
peso, but it was reluctant to do
this because of the effect high
interest rates would have had
on the real economy and the
banking sector. Certainly by
December 1994 this strategy
would have imposed large
costs.

Evaluating the Evidence on NAFTA
and Capital Flows
This section evaluates the evidence on
NAFTA and capital flows to see whether it
is consistent with either of the hypotheses
that NAFTA caused the peso crisis through
its effect on capital flows. The first subsection examines the evidence on the extent
to which NAFTA encouraged capital inflows and the next looks at the argument
that capital flows are inherently destabilizing. Finally, the role of NAFTA in the Chiapas uprising and political instability is appraised.

“On January 1, 1994—the day
that the North American FreeTrade Agreement (NAFTA) took
effect, binding Mexico’s modernizing economy to that of the United
States—Indian peasants at the
southern end of the country rose
in armed rebellion. ... Many in
Chiapas fear that NAFTA will
worsen the existing divide be-

Evidence on NAFTA and Capital Inflows.
Mexico did indeed have low and falling national savings rates—4 percent of GDP in
1994, for example (see Table 1)—making it
more dependent on international capital
flows. Net savings fell from 10.8 percent of

FEDERAL RESERVE BANK

43

OF

S T. L O U I S

J U LY /A U G U S T 1996

18

Data from OECD (1995) National Accounts.

19

See Sachs, Tornell and Velasco
(1995a and 1995b).

GDP in 1990 to 4.1 percent of GDP in
1994.18 This reduction in savings was driven by corresponding increases in private
and government consumption, which rose
2.5 percentage points and 4.1 percentage
points, respectively, over the same period.
There are several problems with the
hypothesis that the declining savings was a
result of NAFTA. First, it is not very plausible that NAFTA would cause a large rise
in private (or government) consumption.
Trade barriers cause consumers to substitute one form or source of consumption
for another but change aggregate levels of
consumption/saving relatively little. Thus
the effect of trade liberalization on trade
deficits is not likely to be very big. Also,
the fact that most of the increase in consumption was caused by a rise in government consumption does not fit well with
the hypothesis that NAFTA caused the fall
in savings. The sluggish economy in 1993
and election year politics in 1994 were
more likely than NAFTA to have played a
role in this relaxation of fiscal policy. Finally, the timing of the inflows is wrong;
the inflows started in 1990 with the resolution of the debt crisis and the liberalization of capital account rules to permit foreigners to hold government bonds and
nonvoting equity shares in Mexican
firms.19 Figure 4 illustrates this rise in the
capital account surplus; the majority of
capital inflows had entered Mexico well
before NAFTA was negotiated, much less
enacted. Other economic reforms, like the
decline in inflation and the privatization of
state-owned industries, also helped drive
investment.
While NAFTA was not the only or
even the major causal factor for capital
inflows, it probably had some marginal
effect. Figure 4 shows that Mexico’s capital account surplus did peak in the first
quarter of 1994, coinciding with the implementation of NAFTA. The figure is
somewhat deceptive in that a surge in
inflows in January 1994 and February
1994 was masked in the quarterly capital
account figures by a major outflow in
March after the assassination of presidential candidate Luis Donaldo Colosio.

Part of the capital inflow was soaked up
in the form of a rapid increase and then
decrease in official reserves—shown by
the spike in Figure 3 at the beginning
of 1994. That is, the Bank of Mexico
bought up dollars in sterilized intervention
to keep the price of the peso down in
January and February 1994. The surge
was not out of proportion to earlier flows,
however.
To the extent that private Mexican
consumption increased in the early 1990s,
there are many factors aside from NAFTA
to explain it. Prolonged slow growth
(since 1980) had created repressed consumer demand. After growth returned in
1988, consumption spending rose along
with it. Also, to the limited extent that reducing trade barriers may change savings
and consumption decisions, NAFTA was
not the only trade initiative. Mexico engaged in unilateral trade liberalization and
trade agreements with Chile, Colombia,
Venezuela, and Costa Rica. Similar to
other developing countries, economic reform and financial liberalization—quite
apart from NAFTA—raised expectations of
increased future income and gave more
Mexicans access to credit.
To summarize: the evidence does
not support the argument that NAFTA
drove large capital inflows to Mexico.
NAFTA did increase foreign confidence
and marginally increased capital inflows,
but most capital inflows had entered before
passage of NAFTA. In fact, NAFTA may
have delayed a crisis by drawing in capital
that supported the peso in early 1994.
Volatility of Capital Flows. The question
of whether capital flows are excessively
volatile or inherently destabilizing is difficult to answer because capital should exit
a country in response to poor economic
policies or other factors that reduce its
productivity. This helps ensure that capital
is as productive as possible and provides
governments with an immediate incentive
to maintain sound policies. On the other
hand, it is possible that portfolio investment overreacts to information, and this
volatility does create problems.

FEDERAL RESERVE BANK

44

OF

S T. L O U I S

J U LY /A U G U S T 1996

Because capital does move rapidly out
of weak currencies in moments of crisis
and these movements can be destabilizing,
some economists have advocated a very
small tax on international financial transactions to deter short-term speculation.20
Trying to eliminate international capital
flows would be a mistake, however, because capital inflows can be quite helpful
in promoting development. Also, they are
not necessarily destabilizing. Rather, their
volatility can depend on the soundness of
macroeconomic policies followed in the
recipient countries. Further, outflows
occur without regard to the nationality of
the investors in the presence of unsound
macroeconomic policies. Domestic residents would get their money out of the
domestic assets under the same conditions
as international investors, if the value of
these assets were threatened.

capital flows provide major advantages for
both investors and recipients.
This movement of assets can also
cause difficulties, however. Corbo and
Hernandez (1996) studied the problems
posed by this movement of assets in nine
countries: Argentina, Chile, Columbia, Indonesia, Malaysia, Mexico, the Philippines, South Korea, and Thailand. They report that though the absolute level of
investment in Mexico from 1986 to 1993
was very large compared with the other
countries, Thailand, Malaysia, and Chile
received larger capital inflows as a percentage of GDP than did Mexico. Many of
these countries have also encountered
problems similar to those confronted by
the Mexican authorities. For example, in
regimes with fixed or predetermined exchange rates, capital inflows can lower domestic interest rates, raise domestic expenditures and temporarily raise inflation,
which can lead to an overvalued currency
and large trade deficits.
Partly to offset the tendency toward
overvaluation caused by the capital flows,
all of these countries have undertaken liberalization of trade, though none of them
has concluded a trade agreement comparable in importance to NAFTA. But free
trade agreements are not necessary to create substantial capital inflows. The breadth
and size of these capital flows to reforming
countries in the developing world in the
last 10 years makes it difficult to believe
that NAFTA was the primary reason for
the inflows to Mexico.
Capital flow volatility poses particular
problems for fixed exchange rate regimes
because capital outflows are synonymous
with exchange rate crises. Investors who
perceive a possibility of a discrete fall in
the value of their assets (that is, a devaluation), will attempt to get their money out
of the weak currency. Thus crises appear
suddenly when capital is easily moved.
These outflows are merely a symptom of
the problem, however, not the cause.

NAFTA and the Chiapas Uprising.
NAFTA may have been a catalyst for, but
certainly was not the cause of, the Chiapas
uprising. This rebellion reflected grievances long and deeply felt by the impoverished south against the more prosperous
north. Also, the uprising was only one
political shock among many that Mexico
endured that year, including two major
assassinations, a rise in U.S. interest rates
and a presidential election. If the December Chiapas uprising had not sparked the
crisis, something else likely would have.
Capital Flows to Emerging Markets. Mexico is not the only developing country to
experience heavy capital inflows recently.
In the last 10 years capital inflows to developing countries have increased sharply
because of two factors: market-oriented
policy reforms and low interest rates prevailing in the developed world. These factors draw in capital because policy reforms
raise the return to investment in developing countries and the low interest rates in
the developed world provide a less attractive alternative for international investors.
For developing countries, capital flows
provide a much needed source of funds for
economic growth. Ideally, international

Summary of the Evidence on NAFTA and
Capital Flows. NAFTA is an unlikely culprit to blame for the quantity of capital in-

FEDERAL RESERVE BANK

45

OF

S T. L O U I S

20

See Frankel (1996) for a short
discussion of the Tobin tax.

J U LY /A U G U S T 1996

flows Mexico received in the early 1990s.
The surge in capital flows started well before the enactment of NAFTA and had
more to do with the rise in consumption
by the Mexican government and the other
economic reforms undertaken.
Whatever the source or timing of the
inflows, however, NAFTA was not responsible for the outflows. It is generally acknowledged that the outflows were generated by some combination of inconsistent
policies and political shocks that generated
a liquidity crisis; the Mexican government
had more short-term obligations—in the
form of dollar-linked bonds—coming due
than it had liquid assets.21

REFERENCES
Aguilar, Linda M. “NAFTA: A Review of the Issues,” Federal Reserve
Bank of Chicago Economic Perspectives (Jan/Feb 1993),
pp. 12–20.
Anderson, Sarah, John Cavanagh and David Ranney, eds. NAFTA’s First
Two Years—The Myths and the Realities. Washington D.C.: The Institute for Policy Studies, 1996.
Brown, Drusilla K., Alan V. Deardorf and Robert M. Stern. “A North
American Free Trade Agreement: Analytical Issues and a Computational Assessment,” The World Economy (January 1992),
pp. 11–29.
Calvo, Guillermo. “Capital Flows and Macroeconomic Management:
Tequila Lessons,” unpublished manuscript (March 1996).
Corbo, Vittorio, and Jaime de Melo. “Lessons from the Southern Cone
Policy Reforms,” World Bank Research Observer (July 1987),
pp. 111–42.

CONCLUSION

21

Because the Bank of Mexico
had ultimate control over the
supply of pesos, it is true, by
definition, that the devaluation
was caused by insufficiently
tight monetary policy. That is,
the Mexican authorities’ exchange rate and growth policy
objectives were mutually inconsistent. For a discussion of the
policy priorities of the Mexican
authorities, see Gruben (1996)
or Gil-Diaz and Carstens
(1995). Overviews of the
events leading to the crisis may
be found in OECD (1995), IMF
(1995) or GAO (1996).
Calvo (1996) and Garber and
Lall (1996) discuss the roles
played by capital flows and
weakness in the Mexican financial system. Sachs, Tornell and
Velasco (1995a and 1995b)
discuss the problems created
by Mexico’s dollar-linked debt
instruments.

_______, and Leonardo Hernandez. “Macroeconomic Adjustment to
Capital Inflows: Lessons from Recent Latin American and East Asian
Experience.” World Bank Research Observer (February 1996), pp.
61–85.

As Mexico entered into NAFTA at the
beginning of 1994, it was widely and correctly applauded as a model of economic reform. Before the end of the year, however, it
was forced to first devalue and later to allow
the peso to float. In early 1995, it was forced
to borrow from the IMF and the United
States to get through a liquidity crisis.
Critics of NAFTA such as Ross Perot,
Pat Buchanan, William Greider, and
Robert Kuttner blamed the trade treaty
for this crisis. This article examines two
versions of this argument: that Mexican
policymakers manipulated the value of
the peso because of NAFTA and that
NAFTA caused volatile international capital flows that brought down the peso.
The evidence does not support the hypothesis that the crisis could have resulted from NAFTA’s economic effects.
Any peso overvaluation in 1994 resulted
from the use of the exchange rate to reduce inflation, a common consequence of
this strategy. Although capital inflows can
present problems and aggravate instability
in developing countries, they are also
very useful to promote economic development. In any case, the flows to Mexico
were only partially driven by NAFTA.
NAFTA was not, in any sense, responsible for the devaluation, but this episode
reminds us that good policies can have
unintended consequences.

Dornbusch, Rudiger, and Alejandro Werner. “Mexico: Stabilization, Reform and No Growth,” Brookings Papers on Economic Activity, vol. 1,
(1994), pp. 253–97 .
Freidman, Milton. “A Theory of the Consumption Function.” Princeton
University Press, 1957.
Garber, Peter M., and Subir Lall. “The Role and Operation of Derivative
Markets in Foreign Exchange Market Crises,” unpublished manuscript
(February 1996).
Gil-Diaz, Francisco, and Agustin Carstens. “One Year of Solitude: Some
Pilgrim Tales About Mexico’s 1994-1995 Crisis,” The American
Economic Review (May 1996), pp. 164–9.
_______, and _______. “Some Hypotheses Related to the Mexican 1994-1995 Crisis,” Banco de Mexico, Serie Documentos de Investigacion (1995) 9601.
Greider, William. “Southern Comfort. How Come There Are Billions to
Bail Out Mexico But Nada for U.S. Cities?” Rolling Stone (March 9,
1995), pp. 40–2.
Gruben, William C. “Policy Priorities and the Mexican Exchange Rate Crisis,” Federal Reserve Bank of Dallas Economic Review (First Quarter
1996), pp. 19–29.
Hufbauer, Gary Clyde, and Jeffrey J. Schott. NAFTA: An Assessment . Institute for International Economics, 1993.
International Monetary Fund. “Evolution of the Mexican Crisis,” in International Capital Markets. Developments, Prospects, and Policy Issues,
IMF (1995), pp. 53–69.
Kehoe, Timothy J. “A Review of Mexico’s Trade Policy from 1982 to
1994,” The World Economy: Global Trade Policy (1995), pp.
135–51.

FEDERAL RESERVE BANK

46

OF

S T. L O U I S

J U LY /A U G U S T 1996

Krugman, Paul. “Review of: NAFTA: An assessment,” Journal of Economic Literature (June 1995), pp. 849–51.
_______. “The Uncomfortable Truth about NAFTA: Its Foreign Policy,
Stupid,” Foreign Affairs (Nov.-Dec. 1993), pp. 13–19.
Kuttner, Robert. “Trouble in Mexico no surprise, misguided NAFTA leads
to drop in peso,” Akron Beacon Journal (January 22, 1995),
p. C1.
“Mexico’s second-class citizens say enough is enough,” The Economist
(January 8, 1994), pp. 41–3.
Organization for Economic Co-operation and Development. OECD Economic Surveys 1994-1995, Mexico (1995).
Oliver, Christian. Conference Summary of “Did NAFTA Kill the Peso?”
American Enterprise Institute for Public Policy Research, January 30,
1996.
Orme, William. “Myths versus Facts: The Whole Truth about the HalfTruths,” Foreign Affairs (Nov.-Dec. 1993), pp. 2–12.
Perot, Ross. “Perspective on NAFTA,” Los Angeles Times (January 4,
1995), p. 7.
Rowen, Hobart. “Administration Ignored Peso Warnings,” The Washington Post (February 5, 1995), p. H2.
Sachs, Jeffrey, Aaron Tornell and Andres Velasco. “The Collapse of the
Mexican Peso: What Have We Learned?” National Bureau of Economic Research Working Paper (June 1995a), 5142.
_______, _______ and _______. “The Real Story,” The International Economy (March/April 1995b), pp. 14–17 and 50–1.
Schott, Jeffrey J. “NAFTA: An American Perspective,” International Trade
Journal (Spring 1994), pp. 3–8.
Tornell, Aaron, and Gerardo Esquivel. “The Political Economy of Mexico’s
Entry to NAFTA,” National Bureau of Economic Research Working
Paper (October 1995), 5322.
United States General Accounting Office (GAO). “Mexico’s Financial Crisis. Origins, Awareness, Assistance, and Initial Efforts to Recover.” Report to the Chairman, Committee on Banking and Financial Services,
House of Representatives, February 1996.
United States Senate. “The Mexican Peso Crisis,” Hearings before the
Committee on Banking, Housing, and Urban Affairs (Jan. 31, 1995,
March 9-10, May 24, 1995, and July 14, 1995).
Velasco, Andres. “Lessons from the Recent Mexican Crisis,” CV Starr
Newsletter (vol. 13, 1995).

FEDERAL RESERVE BANK

47

OF

S T. L O U I S

J U LY /A U G U S T 1996

Apostolos Serletis is a professor of economics at the University of Calgary, and David Krause is a Ph.D. student in the Department of
Economics at the University of Calgary. The authors would like to thank Bill Gavin, Michael Pakko, and Don Allen for comments that
greatly improved this paper.

Nominal
Stylized Facts
of U.S.
Business
Cycles

movements of money and prices, appear to
be business cycle myths. In contrast to
conventional wisdom, they argue that the
price level (whether measured by the implicit GNP deflator or by the consumer
price index), is countercyclical. Although
the monetary base and M1 are both procyclical, neither leads the cycle. This evidence counters Mankiw’s (1989) criticism
of real business cycle models on the
grounds that they do not predict procyclical variation in prices. Moreover, the evidence of countercyclical price behavior has
been confirmed by Cooley and Ohanian
(1991), Backus and Kehoe (1992), Smith
(1992), and Chadha and Prasad (1994).
The cyclical behavior of money and
prices has important implications for the
sources of business cycles and therefore
for discriminating among competing models. Initially it was argued, for example,
that procyclical prices will be consistent
with demand-driven models of the cycle,
whereas countercyclical prices would be
consistent with predictions of supplydetermined models, including real business
cycle models. Subsequently, however, Hall
(1995) has shown that adding more detail
to traditional demand-driven models can
produce countercyclical prices, whereas
Gavin and Kydland (1995) have shown
that alternative money supply rules can
generate either procyclical or countercyclical prices in a real business cycle setting.
The objective of this paper is to reexamine the cyclical behavior of money
and prices using monthly U.S. data. For
comparison purposes, the methodology
used is mainly that of Kydland and
Prescott (1990). Therefore in accordance
with the real business cycle approach to
economic fluctuations, we define the
growth of a variable as its smoothed
trend and the cycle components of a
variable as the deviation of the actual values of the variable from the smoothed
trend. However, we investigate robustness
of the results to alternative (relevant)

Apostolos Serletis and
David Krause

T

his paper investigates the basic nominal stylized facts of business cycles in
the United States using monthly data
from 1960:1 to 1993:4 and the methodology suggested by Kydland and Prescott
(1990). Comparisons are made among
simple-sum and Divisia aggregates using
the Thornton and Yue (1992) series of Divisia monetary aggregates. The robustness
of the results to (relevant) nonstochastic
stationarity-inducing transformations is
also investigated.
Kydland and Prescott (1990) argue
that business cycle research took a wrong
turn when researchers abandoned the effort to account for the cyclical behavior of
aggregate data following Koopmans’s
(1947) criticism of the methodology developed by Burns and Mitchell (1946) as
being “measurement without theory.”
Crediting Lucas (1977) with reviving interest in business cycle research, Kydland
and Prescott initiated a line of research
that builds on the growth theory literature.
Part of it involves an effort to assemble
business cycle facts. This boils down to investigating whether deviations of macroeconomic aggregates from their trends are
correlated with the cycle, and if so, at
what leads and lags.
Kydland and Prescott (1990) report
some original evidence for the U.S. economy and conclude that several accepted
nominal facts, such as the procyclical

FEDERAL RESERVE BANK

49

OF

S T. L O U I S

J U LY /A U G U S T 1996

nonstochastic stationarity-inducing transformations.
To highlight the influence of money
measurement on statistical inference [as in
Belongia (1996)], comparisons are made
among simple-sum and Divisia monetary
aggregates (of M1A, M1, M2, M3, and L)
—see Barnett, Fisher, and Serletis (1992)
regarding the state of the art in monetary
aggregation. The money measures employed are monthly simple-sum and Divisia indexes (from 1960:1 to 1993:4), as
described in Thornton and Yue (1992), and
were obtained from the Federal Reserve
Economic Data (FRED) bulletin board of
the Federal Reserve Bank of St. Louis.
The paper is organized as follows. Section 1 briefly discusses the Hodrick
Prescott (HP) filtering procedure for decomposing time series into long-run and
business cycle components. Section 2 presents HP empirical correlations of money,
prices, and nominal interest rates with
industrial production. In section 3 we investigate the robustness of our results to
alternative stationarity-inducing transformations, and in the last section we summarize the main results and conclude.

age to the quarterly components defined
by µ = 1,600 which is commonly used to
define business cycle fluctuations in research literature.
We measure the degree of co-movement
of a series with the pertinent cyclical variable by the magnitude of the correlation
coefficient r(j), je{0, ±1, ±2, ...}. The contemporaneous correlation coefficient—
r(0)—gives information on the degree of
contemporaneous co-movement between
the series and the pertinent cyclical variable. In particular, if r(0) is positive, zero,
or negative, we say that the series is procyclical, acyclical, or countercyclical, respectively. In fact, for 0.23 ≤ |r(0)| < 1,
0.10 ≤ |r(0)| < 0.23, and 0 ≤ |r(0)| < 0.10,
we say that the series is strongly contemporaneously correlated, weakly contemporaneously correlated, and contemporaneously uncorrelated with the cycle,
respectively. Following Fiorito and
Kollintzas (1994) in our sample of 400 observations, the cutoff point 0.1 is close to
the value 0.097 that is required to reject
the null hypothesis, H0 : r(0) = 0, at the
5 percent level in a two-sided test for bivariate normal random variables. Also, the
cutoff point 0.23 is close to the value of
0.229 that is required to reject the null hypothesis H0: |r(0)| ≤ 0.5, in the corresponding one-tailed test. Also, r(j), je{±1,
±2,...}—the cross correlation coefficient—
gives information on the phase-shift of the
series relative to the cycle. If |r(j)| is maximum for a negative, zero, or positive j, we
say that the series is leading the cycle by j
periods, is synchronous, or is lagging the
cycle by j periods, respectively.

METHODOLOGY
For a description of the stylized facts,
we follow the current practice of detrending the data with the HP filter—see
Prescott (1986). For the logarithm of a
time series Xt, for t = 1,2,...,T, this procedure defines the trend or growth component, denoted τt, for t = 1,2,...,T, as the solution to the following minimization
problem
min

τt

T

T−1

(X − τ ) ^ 3(τ
^
t=1
t=2
t

t

2+µ

t+1

HODRICK-PRESCOTT
STYLIZED FACTS

4

− τt) − (τt − τt−1)

2

In Table 1 we report contemporaneous
correlations, as well as cross correlations
(at lags and leads of one through six
months) between the cyclical components
of money and the cyclical component of
industrial production. We see that all the
monetary aggregates are strongly procyclical. With a minor exception for M1A, for
both Divisia and simple-sum measures, the

so Xt − τt is the filtered series. The larger
the µ, the smoother the trend path, and
when µ = ∞, a linear trend results. In our
computations, we set µ = 129,600, as it
has been suggested for monthly data. Note
that the monthly cyclical components defined by µ = 129,600 approximately aver-

FEDERAL RESERVE BANK

50

OF

S T. L O U I S

J U LY /A U G U S T 1996

Table 1

Correlations of HP-Filtered Sum and Divisia
Monetary Aggregates with Industrial Production*
Correlation Coefficients of Industrial Production with
Variable, x

Volatility

xt26

xt25

xt24

xt23

xt22

xt21

xt

xt11

xt12

xt13

xt14

xt15

xt16

Sum M1A
Sum M1
Sum M2
Sum M3
Sum L
Divisia M1A
Divisia M1
Divisia M2
Divisia M3
Divisia L

2.09
1.93
1.41
1.48
1.11
1.74
1.50
1.81
1.78
1.58

0.43
0.37
0.71
0.50
0.33
0.39
0.28
0.67
0.68
0.62

0.43
0.37
0.70
0.52
0.39
0.40
0.28
0.65
0.67
0.63

0.43
0.37
0.66
0.53
0.44
0.40
0.29
0.62
0.66
0.64

0.43
0.36
0.62
0.53
0.49
0.40
0.28
0.59
0.64
0.64

0.42
0.35
0.56
0.52
0.52
0.39
0.27
0.54
0.60
0.62

0.40
0.32
0.49
0.50
0.55
0.37
0.24
0.47
0.56
0.60

0.38
0.28
0.40
0.47
0.57
0.35
0.21
0.40
0.50
0.57

0.35
0.24
0.32
0.44
0.58
0.32
0.18
0.33
0.45
0.53

0.31
0.19
0.24
0.41
0.59
0.29
0.14
0.25
0.39
0.49

0.28
0.15
0.16
0.38
0.58
0.27
0.10
0.19
0.34
0.45

0.25
0.11
0.09
0.35
0.58
0.26
0.08
0.13
0.29
0.41

0.24
0.08
0.03
0.32
0.56
0.25
0.05
0.08
0.25
0.37

0.22
0.05
20.03
0.29
0.55
0.24
0.03
0.03
0.21
0.33

* Monthly data from sample period 1960:1–1993:4.

broader the aggregate the more procyclical
it is. There is also evidence that M2
money, however defined, leads the cycle by
more than the other aggregates and, if anything, Sum L is slightly lagging. These results suggest the only major differences
among simple-sum and Divisia monetary
aggregates occur in the stronger correlation at leads for the broad Divisia aggregates, M3 and L.
We interpret these results as being
generally consistent with the cyclical
money behavior in the United States reported (using quarterly data) by Kydland
and Prescott (1990) and Belongia (1996).
Unlike Belongia, who like Kydland and
Prescott, uses quarterly data and only
the simple-sum and Divisia measures of
M1 and M2, we find no significant differences across narrow simple-sum and Divisia monetary aggregates. We find strong
contemporaneous correlations between
broad-sum and Divisia money and the
cyclical indicator. Divisia L, however, is
leading the cycle, and Sum L is slightly
lagging the cycle. This result seems to be
consistent with the evidence reported by
Barnett, Offenbacher, and Spindt (1984),
who found that Divisia L was the best aggregate in terms of causality tests, produced the most stable demand-for-money

function, and provided the best reducedform results.
Next we turn to the statistical properties of the cyclical components of the price
level (measured by the consumer price
index) and two short-term nominal interest rates (to deal with anomalies that arise
because of different ways of measuring financial market price information)—the
Treasury bill rate and the commercial
paper rate. The Treasury bill rate is the interest rate on short-term, unsecured borrowing by the U.S. government, whereas
the commercial paper rate is the interest
rate on short-term, unsecured borrowing
by corporations. As Friedman and Kuttner
(1993, p. 194) argue, the commercial
paper rate is superior in capturing the information in financial prices because “the
commercial paper rate more directly reflects the cost of finance corresponding to
potentially interest-sensitive expenditure
flows than does the Treasury bill rate.”
Table 2 reports HP cyclical correlations of prices and short-term nominal interest rates with industrial production. We
see that the price level is strongly countercyclical, whereas both the Treasury bill
rate and the commercial paper rate are
strongly procyclical and lag the cycle.
These results provide strong confirmation

FEDERAL RESERVE BANK

51

OF

S T. L O U I S

J U LY /A U G U S T 1996

Table 2

Correlations of HP-Filtered Prices and Short-Term
Nominal Interest Rates with Industrial Production*
Correlation Coefficients of Industrial Production with
Variable, x

Volatility

xt26

xt25

xt24

xt23

xt22

xt21

xt

xt11

xt12

xt13

xt14

xt15

xt16

Consumer
Price Index
Treasury Bill
Rate
Commercial
Paper Rate

1.46

20.73 20.71 20.68 20.65 20.60 20.55 20.48 20.43 20.37 20.31 20.25 20.20 20.15

1.66

20.17 20.09

0.01

0.11

0.22

0.32

0.40

0.44

0.46

0.47

0.47

0.48

0.48

1.44

20.12 20.03

0.05

0.15

0.25

0.33

0.39

0.42

0.43

0.43

0.43

0.43

0.43

* Monthly data from sample period 1960:1–1993:4.

for the countercyclical price behavior in
the United States reported by Kydland and
Prescott (1990), Cooley and Ohanian
(1991), Backus and Kehoe (1992), Smith
(1992), and Chadha and Prasad (1994).
They clearly support the Kydland and
Prescott (1990) claim that the perceived
fact of procyclical prices is but a myth.

ries, argue that the interpretation of HP
stylized facts depends on assumptions
about the time series properties of the
original data. For example, when the original data are trend stationary, the HP filter
operates like a high-pass filter. That is, it
removes the low frequency components
and allows the high frequency components to pass through. When the original
data are difference stationary, however,
the HP filter does not operate like a highpass filter. In this case, HP stylized facts
about periodicity and co-movement are
determined primarily by the filter and reveal very little about the dynamic properties of the original data.
More recently, however, Baxter and
King (1995) argue that HP filtering can produce reasonable approximations of an ideal
business cycle filter. Though we believe that
the results based on the HP filter are reasonably robust across business cycle filters, we
believe it is useful to compare what we are
doing with alternative popular methods of
detrending the data. Once, however, we abstract from growth theory, we need to make
some assumption about the trend. In particular, deterministic detrending will be the
appropriate stationarity-inducing transformation under trend stationarity and differencing under difference stationarity.
Results reported in Koustas and
Serletis (1996), based on augmented
Dickey-Fuller–type regressions, indicate
that the null hypothesis of a unit root in
levels cannot be rejected for any of the

ROBUSTNESS TO
STATIONARITY-INDUCING
TRANSFORMATIONS
We have characterized the key nominal features of U.S. business cycles using a
modern counterpart of the methods developed by Burns and Mitchell (1946)—HP
cyclical components. The HP filter is used
almost universally in the real business
cycle research program and extracts a
long-run component from the data, rendering stationary series that are integrated
up to the fourth order. HP filtering, however, has recently been questioned as a
unique method of trend elimination. For
example, King and Rebelo (1993) argue
that HP filtering may seriously change
measures of persistence, variability, and
co-movement. They also give a number of
examples that demonstrate that the dynamics of HP filtered data can differ significantly from the dynamics of differenced
or detrended data.
Also, Cogley and Nason (1995), in
analyzing the effect of HP filtering on
trend- and difference-stationary time se-

FEDERAL RESERVE BANK

52

OF

S T. L O U I S

J U LY /A U G U S T 1996

Table 3

Correlations of First Differences of Sum and Divisia Money
with First Differences of Industrial Production*
Correlation Coefficients of Industrial Production with
Variable, x

Volatility

xt26

xt25

xt24

xt23

xt22

xt21

xt

xt11

xt12

xt13

xt14

xt15

xt16

Sum M1A
Sum M1
Sum M2
Sum M3
Sum L
Divisia M1A
Divisia M1
Divisia M2
Divisia M3
Divisia L

0.005
0.004
0.003
0.003
0.003
0.005
0.004
0.004
0.003
0.003

0.09
0.09
0.25
0.16
0.10
0.04
0.04
0.18
0.17
0.14

0.06
0.08
0.23
0.18
0.11
0.03
0.05
0.20
0.20
0.16

0.05
0.07
0.21
0.17
0.07
0.01
0.03
0.17
0.18
0.13

0.17
0.17
0.27
0.21
0.12
0.14
0.14
0.28
0.27
0.24

0.14
0.14
0.23
0.17
0.10
0.11
0.10
0.25
0.24
0.21

0.12
0.12
0.16
0.13
0.14
0.09
0.08
0.20
0.19
0.21

0.10
0.05
0.11
0.11
0.15
0.04
20.02
0.07
0.08
0.12

0.06
0.04
0.04
0.10
0.17
0.02
20.01
0.02
0.06
0.12

20.04
20.05
20.07
0.04
0.11
20.07
20.06
20.09
20.03
0.02

20.08
20.12
20.07
0.03
0.11
20.08
20.10
20.10
20.04
0.02

20.08
20.08
20.07
0.05
0.11
20.06
20.04
20.07
0.00
0.04

20.03
20.04
20.05
0.04
0.08
20.02
20.01
20.05
0.01
0.05

20.06
20.05
20.04
0.01
0.09
20.02
20.04
20.04
0.02
0.07

*Monthly data from sample period, 1960:1–1993:4.

Table 4

Correlations of First Differences of Prices and Short-Term Nominal Interest Rates
with First Differences of Industrial Production*
Correlation Coefficients of Industrial Production
Variable, x

Volatility

xt26

xt25

xt24

xt23

xt22

xt21

xt

xt11

xt12

xt13

xt14

xt15

xt16

Consumer
Price Index
Treasury Bill
Rate
Commercial
Paper Rate

0.003

20.23

20.22

20.30

20.23

20.23

20.16

20.08

20.07

20.07

0.00

20.01

20.04

20.07

0.006

20.10

20.06

20.06

20.08

0.13

0.20

0.30

0.24

0.18

0.04

20.00

0.02

20.00

0.006

20.03

20.02

20.08

20.04

0.14

0.21

0.23

0.23

0.11

0.03

20.03

0.02

0.02

*Monthly data from sample period 1960:1–1993:4.

series used here, whereas the null hypothesis of a second unit root is rejected
except for Sum M3, Sum L, and the price
level which appear to be integrated of
order 2 [or I(2) in Engle and Granger
(1987) terminology]. Based on this evidence, in Tables 3 and 4 we report correlations (in the same fashion as in Tables
1 and 2) based on differenced data, keeping in mind that although differencing
yields stationary series, these stationary
series do not in general correspond to
cyclical components. See, for example,
Baxter and King (1995). These results are
generally supportive of the hypothesis of

acyclical money and price behavior. Nominal interest rates appear to be strongly
procyclical and lagging slightly.

CONCLUSION
In this paper we investigated the cyclical behavior of U.S. money, prices, and
short-term nominal interest rates, using
monthly data from 1960:1 to 1993:4 and
the methodology of Kydland and Prescott
(1990). Based on stationary HP cyclical
deviations, our results fully match recent
evidence on the countercyclicality of the
price level. We also found that short-term

FEDERAL RESERVE BANK

53

OF

S T. L O U I S

J U LY /A U G U S T 1996

Koopmans, Tjalling. “Measurement Without Theory,” Review of Economics and Statistics (1947), pp. 161–72.

nominal interest rates are strongly procyclical and that money is in general procyclical. Furthermore, the evidence suggests that there are only slight differences
across narrow simple-sum and Divisia
money measures.

Koustas, Z. and A. Serletis. “Monetary Aggregation and the Quantity Theory of Money.” Unpublished manuscript, The University of Calgary
(1996).
Kydland, Finn E. and Edward C. Prescott. “Business Cycles: Real Facts
and a Monetary Myth,” Federal Reserve Bank of Minneapolis Quarterly
Review (Spring 1990), pp. 3–18.

REFERENCES

________ and _______. “Time to Build and Aggregate Fluctuations,” Econometrica (1982), pp. 1345–70.

Backus, David K., and Patrick J. Kehoe. “International Evidence on the
Historical Properties of Business Cycles,” The American Economic Review (1992), pp. 864–88.

Lucas, Robert E. Jr. “Understanding Business Cycles.” in Karl Brunner and
Allan H. Meltzer, eds., Stabilization of the Domestic and International
Economy, vol. 5 of the Carnegie-Rochester Conference Series on Public
Policy, 1977, pp. 7–29.

Barnett, William A., Douglas Fisher, and Apostolos Serletis. “Consumer
Theory and the Demand for Money,” Journal of Economic Literature
(1992), pp. 2086–119.

Mankiw, N. Gregory. “Real Business Cycles: A New Keynesian Perspective,” Journal of Economic Perspectives (1989), pp. 79–90.

Barnett, W.A., E.K. Offenbacher, and P.A. Spindt. “The New Divisia Monetary Aggregates,” Journal of Political Economy (1984), pp. 1049–85.

Prescott, Edward C. “Theory Ahead of Business Cycle Measurement,”
Federal Reserve Bank of Minneapolis Quarterly Review (1986),
pp. 9–22.

Baxter, Marianne, and Robert G. King. “Approximate Band-Pass Filters
for Economic Time Series,” NBER Working Paper No. 5052 (1995).
Belongia, Michael T. “Measurement Matters: Recent Results from Monetary Economics Re-examined,” Journal of Political Economy (October
1996), pp. 1065–83.

Smith, R. Todd. “The Cyclical Behavior of Prices,” Journal of Money,
Credit, and Banking (1992), pp. 413–30.
Thornton, D.L. and Yue, P. “An Extended Series of Divisia Monetary Aggregates,” The Federal Reserve Bank of St. Louis Review (1992),
pp. 35–52.

Burns, Arthur, and Wesley C. Mitchell. Measuring Business Cycles. NBER
1946.
Chadha, Bankim, and Eswar Prasad. “Are Prices Countercyclical? Evidence from the G7,” Journal of Monetary Economics (1994),
pp. 239–57.
Cogley, Timothy, and James M. Nason. “Effects of the Hodrick-Prescott
Filter on Trend and Difference Stationary Time Series: Implications for
Business Cycle Research,” Journal of Economic Dynamics and Control
(1995), pp. 253–78.
Cooley, T.F., and L.E. Ohanian. “The Cyclical Behavior of Prices,” Journal
of Monetary Economics (1991), pp. 25–60.
Engle, Robert F., and Clive W. Granger. “Cointegration and Error Correction:
Representation, Estimation and Testing,” Econometrica (1987), pp.
251–76.
Fiorito, R., and T. Kollintzas. “Stylized Facts of Business Cycles in the
G7 from a Real Business Cycles Perspective,” European Economic Review (1994), pp. 235–69.
Gavin, William T. and Finn E. Kydland. “Endogenous Money Supply and
the Business Cycle.” Discussion Paper 95-010A, Federal Reserve Bank
of St. Louis (July 1995).
Hall, Thomas E. “Price Cyclicality in the Natural Rate - Nominal Demand
Shock Model,” Journal of Macroeconomics (1995), pp. 257–72.
King, Robert G. and Sergio T. Rebelo. “Low Frequency Filtering and Real
Business Cycles,” Journal of Economic Dynamics and Control (1993),
pp. 207–31.

FEDERAL RESERVE BANK

54

OF

S T. L O U I S