Michael T. Belongia

Michael 71 Belongia is an assistant vice president at theFederal Reserve Bank of St. Louis. Lynn D, Dietrich providedresearch assistance.

Foreign Exchange Interventionby the United States:A Review and Assessment of1985-89

rEnt:p..tHE FEDERAL RESERVE Bank of New York,

acting as an agent of the U.S. Treasury, occa-sionally intervenes in foreign exchange markets.These actions, which involve the purchase orsale of assets denominated in foreign currenciesin exchange for dollars, are intended to affectthe exchange rate by altering the asset suppliesdenominated in one currency relative to that ofanother, Intervention can he directed to a varie-ty of objectives including a change in the levelof the dollar exchange rate, a reduction in thevolatility of the dollar’s value around some level,or an adjustment of the Federal Reserve’s hold-ings of assets denominated in foreign currenciesrelative to its holdings of dollar-denominatedassets.

This article explains the mechanics of foreignexchange intervention for people who are not

s1ecialists in this area of study and identifiesavenues through which sterilized interventionconceivably could affect the exchange rate, Itthen describes and analyzes daily data recently

released by the Federal Reserve Board on its in-tervention activities in the Deutsche mark (DPI)and Japanese yen for the period 1985-89. Be-cause these data had been confidential and simi-lar data from the Bundeshank and Bank ofJapan still are not publicly available, investigat-ing the effects of intervention on the exchangerate has been difficult for at least two reasons,First, without hat-cl data on the activities of allcentral banks involved, statistical tests and infer-ences require assumptions about the unknownactions of some participants. Second, even withcomplete data, stating an hypothesis about theeffectiveness of intervention is subject to fur-ther assumptions about the unknown objectivesof the central banks involved.

Recognizing these limitations, the last sectionof this paper nonetheless attempts to test sever-al hypotheses about the effectiveness of interven-tion.’ While offering no firm conclusions on itseffectiveness during the period examined, the

1See Almekinders and Eijffinger (1991) or Weber (1986) fora survey of issues and evidence on the effectiveness of in-tervention,

newly-released Fed data and some qualitativedata on the timing of foreign central bank inter-ventions give a flavor of the frequency and scaleof these activities, And, under certain assump-tions, these data also identify some circumstancesunder which intervention had statistically sig-nificant effects on the DM/dollar and yen/dollarexchange rates,

An exchange rate is simply the jelative priceof two currencies, So, for example, if twoDeutsche marks can be exchanged for $1, thedollar price of 1DM is $50; conversely, the DPIprice of $1 is 2DM. Just like any relative pricein the marketplace, the value of the exchangerate is determined by the interaction of the sup-ply and demand for the two currencies,

One reason for intervention is to avoid ex-change rates that are higher (or lower) thansome perceived “correct” level, which can havedeleterious short-run effects on a country’s in-ternational trade.~To see how foreign exchangeintervention might be used to affect the level ofthe exchange rate, refer to figure i. panel Ashows the markets for the dollar and DM, usingan exchange rate of 2DM per $1 (1DM 8.50)as an initial market equilibrium. This equilibri-um is shown by points A and A’ at the intersec-tions of the respective supply and demandcurves in the two markets,

Without developing a theoretical model of thevariables that may cause a change in either thesupply or demand for either currency—and,hence, cause a change in the exchange rate—assume that the demand for dollars by Germancitizens rises, shifting D~rightward to D~and

creating a new equilibrium in the market fordollars at point B.3 Here, the new DM price of$1 has risen to 2,5, which implies that, for anequilibrium to exist in the market for DM, thedollar price of one DM must have fallen to 8.40(1/2,5). A change in the equilibrium price of dol-lars will cause a corresponding change in theequilibrium price for DM because changes insupply (demand) of one currency necessarilywill cause corresponding changes in the de-mand (supply) of the other currency to reachthe new equilibrium price in that market. Inother words, if the increased German demandfor dollars raises its price to 2.5 DM, then Ger-man citizens have increased the supply of DMthey are willing to offer for sale in the marketfor any given value of the exchange rate, In-deed, the DM supply curve (S%~,)must shift out-ward until, at its new position (~h~,1),the dollarprice of DM has fallen to the new equilibriumvalue of 8.40,

If this change in the exchange rate weredeemed undesirable by policymakers, then thetwo central banks that control the supplies ofdollars and DPI could alter these supplies to re-store the original exchange rate of 2DM per dol-lar, Though central banks often attempt to actjointly, each reinforcing the action of the other,assume for present purposes that the FederalReserve embarks on intervention alone.~If theFed’s intention is to restoi-e the initial equilibri-um, the figure shows that it must increase thesupply of dollars to a position at S~so that theintersection with D~occurs at the originalequilibrium exchange rate of 2DM per dollar(point C). The Fed would try to accomplish thisby purchasing DM in the open market in ex-change for dollars which, by withdrawing DMfrom circulation, would move the DPI supply

2Purchasing a foreign good requires the exchange of thehome currency for an amount of foreign currency equal tothe foreign price of the good, so that changes in the ex-change rate can affect the amount of goods one countryexports and another imports. Indeed, many observers whoperceive the persistent U.S. deficits in merchandise tradeas a problem have recommended policies to reduce thedollar’s value, Although a reduction in the dollar’s nominalvalue is potentially a short-run stimulant to exports, onlychanges in its real value (nominal value adjusted for pricelevel differences across countries) will have a permanenteffect on exports. See Batten and Belongia (1984) for fur-ther discussion of the distinction between real and nominalexchange rates and the consequences of this distinction indebates about exchange rates and trade flows,

3Throughout, we will focus on the effects of changes inrelative money supplies alone as the main determinants ofexchange rate changes. This is not in itself controversial in

theoretical discussions of exchange rate determination,although economists have been unable to agree on otherelements of a theoretical model to represent exchange ratebehavior, Other factors typically incorporated in suchmodels are differentials between domestic and foreign realgrowth and between domestic and foreign interest rates.This potential role for interest rate differentials is discussedbriefly in the appendix. For discussions of models of ex-change rate determination see Krueger (1983), For a sur-vey of some empirical issues, see Mussa (1979).

4This assumption is made for ease of exposition. Balbach(1978) presents an extensive list of ways that interventionmay be conducted by involving the U,S, Treasury’s Ex-change Stabilization Fund, the Federal Reserve and for-eign central banks. For a specific example of how theBundesbank might act to support the value of the DM, seeBatten and Ott (1984).

34

Figure 1Panel AChanging the Level of the Dollars Value

PM Priceof $11

2.5DM

2PM

Figure 1Panel BStabilizing

DM Price

of $1

25DM

2DM

Dellar Priceef PM

the Dollar’s Value Around a Target Level

PM Priceof $1

2.5DM

2DM

0 1SuM SuM

$30

$-40

Qo Qi Qo

B

QoQI QOM

0 1Ss Ss

Qs

curve back to its original position at S~.In-creasing the quantity of dollars in circulationrelative to PM will reduce the dollar’s value interms of PM. All other factors the same, thissimple example shows how, by altering the rela-tive quantities of any two currencies in circula-tion, central banks might be able to change theprevailing level of the exchange rate. Preciselyhow the Fed would conduct this operation isdiscussed in a later section.

Another point, which will be developed morefully, needs to be mentioned here. As the exam-ple shows, the exchange rate changes becausethe supply of dollars has changed relative to thesupply of PM. Direct intervention is not neces-sary, however, to achieve this result. Instead, acentral bank can affect the exchange rate (bydesign or as a side effect of actions directed toother goals) by changing its money supply dur-ing normal domestic open market operationsand without any sales or purchases of foreign-denominated assets. Conversely, some centralbanks, such as the Bundesbank and Swiss Na-tional Bank, conduct their domestic monetarypolicy through transactions in the foreign ex-change market rather than through transactionsin a domestic credit market in the manner ofthe Fed’s purchases and sales of US. Treasurysecurities,~Therefore, if intervention is to beviewed as an independent tool of monetary poli-cy, it must be able to alter the exchange ratewithout affecting the operations of domesticmonetary policy.

lrfle.rt entio-.n ant.! . I1~Y :tI.rlgt: ~

In addition to altering the level of the ex-change rate, another reason for intervention isto dampen the volatility of exchange rate move-ments. Though the level of the exchange ratemight not change much over time, some peoplebelieve that erratic short-run fluctuations in theexchange rate can be destabilizing. For example,people in two countries may have contracts tobuy or sell a good or service on an agreed date

at an agreed price. A random fluctuation in theexchange rate just before the contract date willinflict unexpected losses on the buyer whosecurrency has depreciated (because he must giveup more units of his domestic currency to getthe same number of units of the foreign curren-cy to pay the seller of the good)- Conversely,the buyer whose currency has appreciated willrealize windfall gains. To the extent these ef-fects occur and are thought to be unpredicta-ble, the exchange rate fluctuations behind themmay impede international trade as exportersand importers perceive such transactions to berisky.°Indeed, the perception that exchangerate volatility impedes trade flows is the ration-ale for the current European Monetary System(EMS) and for the planned move to a single cur-rency in the European Community (EC) by11199.’

To see how intervention can reduce or elimi-nate volatility, consider panel B of figure 1. Forsimplicity, both panels refer only to the marketfor dollars- Referring to the left panel, assumethere are random movements in the demandfor dollars between positions at D~and D~. Fora given supply of dollars, the PM/$ exchangerate will fluctuate between 2.0 and 2.5. To off-set these fluctuations, as shown in the rightpanel, the Federal Reserve could intervene to in-crease the supply of dollars to S~when the de-mand for dollars rises to D~, and intervene toreduce the supply of dollars to S~when the de-mand for dollars falls to D~.Abstracting fromreal-world problems such as adjustment lags,lack of information, changes in the supply ordemand of DM, and other factors that may in-hibit exchange rate smoothing, such a strategyconceivably could keep the DM18 exchange rateat a value of 2.0 as it moved between the sup-ply and demand equilibria represented bypoints A and C. The equilibrium at point B,where the exchange rate rises to 2.5, would beeliminated by successful intervention. Presuma-bly, trade would increase if exporters and im-porters became convinced that interventioneliminated the risk of an exchange rate change.

5The foreign exchange market is used in these cases be-cause the government securities market does not have thedepth necessary for central bank open market operations.

°Buyersand sellers can protect themselves from these loss-es by hedging their transactions; see Williams (1986) formore detail on how hedging of foreign exchange risk canbe accomplished. Rolnick and Weber (1989) report that thecosts of hedging against this sort of exchange rate volatilityrange from 0.5 percent to 3 percent of total foreign sales.

For the U.S. in 1989, this would have put the costs of hedg-ing exchange rate risk between $6.5 billion and $39 billion.

‘The EMS was founded in 1979 to keep bilateral exchangerates among EC currencies within relatively narrowbands—(+) or (—) 2 114 percent of an agreed level wastypical for most currencies. For an overview of the mechan-ics of this system, see Ungereç et al. (1986); for a critique,see Belongia (1988) or Meltzer (1990).

~:iJQ’f’~t:flH~f

In the two examples of intervention just dis-cussed, the value of the exchange rate was al-tered only if there was a change in one moneysupply relative to the other. Moreover, these ex-amples assumed that only the Federal Reservewas intervening. Complications would arise if,on the one hand, one or more other centralbanks reinforced the Fed’s activities by sellingdollars and purchasing DM while the Fed wastrying to reduce the dollar’s value against DM,On the other hand, other central banks couldhave subverted the Fed’s intervention goals hadthey made equal and offsetting purchases ofdollars and sales of DM. Generally speaking, inthe presence of large and highly developedworld markets for the major currencies, the ex-change rate between any two ultimately de-pends on the world supplies and demands forthem rather than on the limited actions of justone central bank.

Although the case of two central banks work-ing at cross-purposes may seem unlikely, in-dividual central banks often take two, largelyoffsetting, actions when they intervene in ex-change markets. This “sterilized” interventionoccurs when a central bank undertakes an openmarket operation in its domestic market that ex-actly offsets the effects of its actions in the for-eign currency market on the domestic moneysupply. Such intervention is called sterilized be-cause the two actions, on net, produce nochange in the domestic money supply. Theproblem, of course, is that, without affectingthe money supply, there is presumably noavenue for sterilized intervention to affect theexchange rate.8 Understanding these procedureswill be important to the statistical tests that fol-low because Federal Reserve interventions areroutinely sterilized.°

To understand sterilized intervention and themechanics behind the shifts in the supply ofdollars shown in figures tA and 1B, considerthe simplified Federal Reserve balance sheet

shown in figure 2. Before any intervention oc-curs, the Fed’s assets are U.S. Treasury securi-ties and its liabilities are the reserves of the U.S.banking system; these two items are the sourcesand uses of the monetary base which, in turn,is the basis of the U.S. money supply. If the Fedwanted to increase the U.S. money supply, itwould make an open market purchase of Treas-ury securities from U.S. banks and pay forthem by crediting the reserve balances of thesebanks at the Fed. Thus, both the assets and lia-bilities sides of the Fed’s balance sheet would in-crease. A simplified Federal Reserve balancesheet showing the effects of injecting reservesinto the US, banking system is depicted in thetop panel of figure 2.

Now consider what happens when the Fed de-cides to engage in foreign exchange interven-tion. Say that at 4:30 am. New York time (10:30am, in Frankfurt, Germany), the Fed and theBundesbank agree that they should try toreduce the value of the dollar by some amount.To do so, the U.S. money supply must be in-creased relative to the German money supply.As a matter of practice, the Fed could purchaseDM-denominated deposits that large U.S. bankshold w-ith German banks and pay for them inthe same way the Fed would conduct a normalopen-market operation: by crediting the reserveaccounts of the U.S. banks that sold their DMdeposits. These transactions are shown in thetwo left-hand accounts in the lower panel offigure 2.

The process does not end there, however.When the drafts made against the PM accountsof U.S. banks are presented by the Fed to theBundesbank for clearing, this transaction addsthe DM deposits of the Fed to the Bundesbank’sliabilities but reduces the reserves of the Ger-man banking system; this is shown in the right-hand columns in the lower panel of figure 2.This draining decline in reserves reduces theGerman money supply. Conversely, the U.S.money supply rises because this transaction in-creases both the assets and liabilities sides of the

°Aquestion arises, in the context of the simple supply anddemand mechanics of exchange rate determination, why acentral bank would engage in sterilized intervention. Twoarguments have been advanced along these lines. A re-cent one, with some empirical support, is that such activi-ties provide a valuable “signal” to participants in theforeign exchange market about the future course of mone-tary policy and its likely effects on future values of the ex-change rate; see Dominguez (1988, 1990). Another

argument, for which the empirical support has been mixed,is that foreign and domestic assets are imperfect substi-tutes and that, by altering their relative supplies throughsterilized intervention, the exchange rate can be changedby affecting the differential between domestic and foreigninterest rates. For a general review of theory and evidence,see Henderson (1984).9See Balbach (1978).

Figure 2

Domestic Open Market Operations and Foreign Exchange Intervention

Panel A: An Expansionary U.S Open Market Operation

Federal Reserve U.S. CommercialBanks (FRB) Banks (cb)

Assets Liabilities Assets Liabilities

~US T’easu-’g -.-Ros~n.es — U.S. Ttasjry

~ecytes of cb secJ’itesReserves

Panel B. U.S. Intervention to Reduce the Dollar

Federal Reserve U.S. Commercial German Commercial

Banks (FRB) Banks (cb) Bundesbank (B) Banks (Gcb)Assets Liabilities Assets Liabilities Assets Liabilities Assets Liabilities

4DM deposits — Reserves — Ros~’ves DM depcsits — Rese’ves — DM depcs’tsatB o~cb cfFAB ofcu

- DM deocs ts — Reservesat Gcb 0~Gcc

Fed’s balance sheet. If no further action istaken, these actions would reflect unsterilizedintervention because they alter the relative sup-plies of dollars and DM.

Under current practices, however, the actionmade at 4:30 am, New York time will bereversed at 11:30 am, when the Open MarketDesk of the New York Fed conducts its domesticopen market operation for the day. To achieveits domestic reserve objectives, the Open MarketDesk assembles projections each day of elementsof the Fed’s balance sheet. Thus, the staff willnote that the foreign exchange activity of sever-al hours ago had the effect of increasing thereserves of the U.S. banking system and will bereflected as an increase in the monetary base,

Consider, for example, that the Open MarketDesk of the New York Fed would have deter-mined—absent any intervention—that the Sys-tem’s domestic objectives for that day would bemet without an open market operation. If theintervention activity caused a $1 billion increasein the monetary base, the domestic Desk, notingthis effect, would undertake a $1 billion sale ofU.S. Treasury securities to reduce reserves.

Thus, the desire to achieve its domestic mone-tary objective has led to actions that cancel thedomestic effects on the money supply of theearlier intervention. Indeed, because offsettingdomestic open market operations leave bankreserves, the monetary base, and, hence, thesupply of dollars unchanged, the only effect ofsterilized intervention is a change in the compo-sition of the Fed’s balance sheet. Its holdings offoreign-denominated assets increase and itsholdings of U.S. ‘treasury securities fall, but thedomestic money supply is unchanged. (See theappendix for a more detailed discussion of howother channels might operate and how sterilizedintervention, in fact, may not be “sterilized” inthe conventional use of the term.)

These mechanics highlight the fact that, if acentral bank wishes to change the nominalvalue of its currency, it need not intervene inthe foreign exchange market at all. Instead, be-cause relative money supplies determine to alarge extent the relative values of the two cur-rencies in question, domestic open market oper-ations alone can have the same effect on theexchange rate as unsterilized intervention.1o Ac-

10Weber (1986) makes this point in much greater detail.

cording to the simple theory outlined in thepreceding section, a central bank wishing toreduce its currency’s nominal value need onlyengineer a rate of money growth faster thanthe growth rate of the money supply in theother country. In this context, the very rapidgrowth rate of Ml in the United States fromMay 1985 through December 1986 has been in-terpreted by some economists as an attempt bythe Federal Reserve to reduce the dollar’s valuewithout engaging in any substantial interventionactivity.” Conversely, an increase in the dollar’svalue can be achieved by engineering a relative-ly slower growth rate of the money supply.12

This suggests that exchange market interventionmay be motivated by a central bank’s desire togive signals to market participants or to alterthe relative shares of domestic and foreign as-sets on its balance sheet, rather than to alterthe level of the exchange rate directly.

,

( ~ C

Figures 3-6 provide a general reference tomovements in the PM/S and yen/S exchangerates and the scale of U.S. intervention duringthe period 1985-89; neither the Fed nor the U.S.Treasury undertook any intervention in 1986.Negative values in the bottom portions of thesefigures indicate a sale of dollars—that is, a pur-chase of the foreign currency by the U.S.authorities. Although matching exchange rateand intervention data in this manner offers thetemptation of making inferences about causeand effect, the discussion that follows will indi-cate that this strategy is not warranted,”

To put figures 3A and 3B, showing exchangerate and intervention data for 1985 in the propercontext, note that the dollai’s value peakedagainst both the yen and PM in February, thenfell by more than 6.5 percent against both cur-rencies by the end of August. Dut-ing theweekend of September 22, 1985, the now-famous Plaza Accord was agreed upon in ameeting of Finance Ministers of the G-5 coun-tries at the Plaza Hotel in New York. In effect,

this agreement pledged support for coordinatedintervention to reduce the dollar’s value.”

In contrast to the press coverage of the time,the data show, first, that declines in the dollar’svalue between February and September occurredwith little or no intervention by the FederalReserve or the Treasury. Figures 3A and 3B alsoshow that the United States limited its interven-tion activities to a period of 34 days immediate-ly following the Plaza agreement. Finally, U.S.cumulative purchases of yen and PM during this34-day period amounted to only $1.44 billionand $1.86 billion (equivalent), respectively.‘these figures can be contrasted with dailyvolume in the New York market alone thataveraged $129 billion per day in April 1989.”‘I’herefore, even if the comparable tradingvolumes of the London and Tokyo markets areignored, (which would raise total daily volumeto in excess of $400 billion) the actual scale ofintervention typically was a trivial share of totalvolume in the foreign exchange market duringthis period.

The data for 1987 in figures 4A and 4B alsohighlight at least two interesting features of re-cent U.S. interventions. The first is that thedominant activity switched from purchasing for-eign currencies and selling dollars—whichwould tend to reduce the dollar’s value—to sell-ing foreign exchange and buying dollars—actions consistent with supporting the dollar’svalue, The figures also show that this interven-tion occurred over a period of steady declinesin the dollar’s value. Indeed, between the timeof the Plaza Accord in September 1985 andMarch 1987, when a new burst of interventionoccurred, the dollar’s value fell from 231.90 yenand 2.73 PM to 151.70 yen and 1.83 PM.Although no rationale was made public for eachintervention, the data for early 1987 are consis-tent with the view that U.S. Treasury officialsbelieved further declines in the value of the do)-lar below these levels should be r-esisted,

The Louvre Accord, reached on February 20,1987, marks a second interesting period. Thisagreement brought together the G-7 Finance

‘‘See, for example, Bernanke and Mishkin (1991).‘2Bordo and Schwartz (1990, pp. 5-6) provide evidence on

relative rates of Ml growth in the United States, Germanyand Japan and corresponding movements in the DM18 andyen/S exchange rates since 1985.

“Briefly, if intervention is directed to resisting exchange ratechanges, the coincidence of a large intervention and a

change in the exchange rate would be evidence on thefailure of intervention.

14For a critique of intervention since the Plaza Accord specif-ically, see Bordo and Schwartz (1990).

“See Federal Reserve Bank of New York (1989).

/

Figure 3aDM/5 Exchange Rate Data for 1985DM/$

3.50

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Figure 3bYen/S Exchange Rate Data for 1985Yen/s

265

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

DM4

Figure 4aDM/$ Exchange Rate Data for 1987

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Figure 4bYen/S Exchange Rate Data for 1987Vents

155

150

145

140

135

130

125

120

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

C / C / / C: C /C /

Figure 5a

DMI$ Exchange Rate Data for 1988

1.15

Jan Feb Mar Apr May Jun Jul Aug Sep

Figure 5bYen/$ Exchange Rate Data for 1988Yen/S140

135

130

125

120

Oct Nov Dec

DM41,95

1 .90

1 .85

1.60

1.70

1.65

1 .60

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Figure 6aDM/$ Exchange Rate Data for 1989DM152.1

2.0

1 .9

4.8

1 .7

Figure 6bYen/S Exchange Rate Data for 1989Vents455

150

145

140

135

130

125

Jan Feb Mar Apr May Jun Jut Aug Sep Oct Nov Dac

CCCCC

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Ministers to agree on intervention activities but,in contrast to Plaza, the Louvre Accord directedtheir attention to stabilizing exchange ratesaround then-current levels rather than changingthe level itself.’°As with Plaza, the UnitedStates followed this agreement with an initialburst of activity in March, the month followingthe agreement; during a period of 53 days, theU.S. authorities sold an equivalent of $4,088 mil-lion in yen and $782 million in PM. The UnitedStates did not intervene again in any concertedway until December when the dollar’s valuewas approaching record lows for the postwarera. Even in this case, however, intervention bythe United States had ceased by January 1988.One is left to speculate, then, whether the ex-change rate levels at this point were consistentwith the target zones established at the Louvremeeting in February or whether the UnitedStates had decided for some other reason tolimit its intervention activities.1

7 The remainingdata for 1988, shown in figures 5A and SB, rev-eal several episodes of intervention on bothsides of the market: selling foreign exchange fordollars in the spring and fall and buying PMagainst dollars in the summer.

The data for 1989 (figures GA and GB), in con-trast to the previous figures, show relativelypersistent and consistent intervention againstboth the yen and PM. For the year as a whole,the United States bought, cumulatively,$10,925.80 million and $11,130.50 million(equivalent) of yen and PM, respectively; in-deed, the United States purchased more PMand yen in 1989 than it did in the years 1985-88combined. At the same time, the dollar’s valuemoved in different directions against the twocurrencies, rising from 123.60 to 143.80 againstthe yen (+ 16 percent) and falling from 1.76 to1.69 against PM (—4 percent). Although it is in~-possible to prove the counterfactual argumentthat the dollar’s value would have changed evenmore without this intervention, its efficacy maybe questioned if intervening in the same man-ner against two different currencies is associat-ed with appreciations against one and deprecia-tions against the other. Certainly this divergingpattern suggests the influence of some otherfactor on the exchange rate beyond intervention.

As noted in the introduction, the story of in-tervention is incomplete without looking at com-parable data on the activities of the Bundesbankand Bank of Japan. Although data on the valueof foreign central bank actions are not available,it is possible to identify days on which foreigncentral banks intervened. In making partitionsof the data set, it is assumed that, at a mini-mum, these central banks were not workingagainst Federal Reserve actions. In other words,we assume that on the days the Fed intervened,the Bundesbank and Bank of Japan either didnothing or they conducted a complementary in-tervention that would reinforce the effect of theFed’s actions.

It also should be noted that it is impossible, orat least very difficult, to refute the counterfac-tual argument—that the exchange rate wouldhave changed without intervention, Indeed,without knowing the reasons for central bankactions, intervention may be motivated primar-ily by a desire to resist—rather than cause—changes in the exchange rate. Because this pos-sibility is a hypothesis that cannot be tested, wewill assume that, in the absence of intervention,the exchange rate would not have changed. Wealso will assume that the purpose of intec’ven-tion was to make the exchange rate move in acertain direction even though a large change onthe day of intervention might indicate the failux-eof an action that was intended to prevent achange in the exchange rate.

In view of the earlier discussion and thesetwo assumptions, several hypotheses merit test-ing. ‘I’he first is that the average absolute changein the exchange rate on days when the Fed in-tervenes alone should not be different from thechanges in the exchange rate when no interven-tion occurs. This test is based on the earlier dis-cussion of sterilized intervention, which predictedthat its failure to change relative money sup-plies also should leave the exchange rate un-affected. A second test is motivated by thepractice of the Bundeshank not to sterilize (atleast not completely) its intervention; thus, on

‘°Funabashi,(1988).‘‘Target zones establish upper and lower bounds for the ex-

change rate. Typically, the exchange rate is tree to varywithin the established range but central banks pledge to

engage in coordinated intervention if the exchange ratebegins to move outside of the range.

Table 1

Mean Changes in the Absolute Value of the DM/$ ExchangeRate (*100)Year X1 (n) X2 (n) X3 (n) X4 (n)

1985 1.00(7) 136(32) 305(12) 205(188)1986 --- 124(15) -- - 1.25(226)1987 1 21(10) 1.01 (19) 1.27 (20) 0.77(194)1988 0.90 (4) 0.87 (42) 1 43 (30) 0.67 (185)1989 076 (20) 084(11) 099 (33) 091 (61)All years o 92(41) 1 07(119) 1.45 (95) 118(834)

Test Statistics for Equality of Mean Changes (Absolute Values)

Year x1=i4 ; *~=i~ i~—*~ ;=i~ x2=x31985 1 59 2 19* t75 0.86 1 41 2.39*1986 - 0.04 - - - . - - - --

1987 175 1.26 284* 067 016 0.871988 069 176 532* 0.07 091 2M21989 075 028 0.49 028 097 0.52All years 128 0.95 1.89 092 1.88 214*

Subscripts to Xs and nufi hypotheses indicate the following= tnterventton only by Federal Reserve

X2 = intervention only by BundeSbank— Intervention by both the Federal Reserve and Bundesbank

no intervention by either central bank* = Statistically significant at the 5 percent level

days when a foreign central bank intervenes rate changes.” On one hand, if intervention isalone, the average absolute change in the cx- associated with larger average daily changes,change rate should be significantly larger than the variance of these changes might rise ason days when no intervention occurs.” The rca- well. On the other hand, if the purpose of inter-son, of course, is that unsterilized intervention vention is to reduce daily volatility in the cx-will affect relative money supplies and, there. change rate and it is successful, the variancefore, should affect the exchange rate as well, could be smaller with intervention, With this

- uncertainty, the test results merely are report-Making clear predictions about the effects of . -

ed, and one can simply make judgments aboutintervention on the volatility of daily exchange .

- . . the effects (if any) of intervention on the van-rate changes, however, is more difficult. Using . *

ance of daily exchange rate changes.the variance of datly absolute changes in the ex-change rate as our measure of volatility, it is The data for means and variances in tables 1-4possible to argue that intervention, conceivably, divide the data into four groups: days when thecould raise or lower the variance of exchange Fed intervened alone, days when a foieign cen-

~tForevidence on the Bundesbank’s practice of not com-pletely sterilizing its intervention, see Neumann (1984).

“Because we are dealing with absolute changes,variance is measured as ~zba’, where n is the number

of observations and e is the exchange rate.

Table 2Mean Changes in the Absolute Value of the Yen/S ExchangeRate (~100)

Year X1 (n) ; (n) X3 (n) X.4 (a)

1985 15056(9) 9789(18) 19273(11) 8810(201)1986 - 97.81 (47) - 86.14 (194)1987 12450(6) 5688(33) 9765(37) 6332(167)1988 7350 (4) 72 12 (17) 62.31 (16) 54.03 (204)1989 9457 (7) 8284 (6) 103.52 (27) 51.48 (85)All years 11782(26) 8231(121) 10467(91) 7097(851)

Test Statistics for Equality of Mean Changes (Absolute Values)

Year ~11X4 2—Xs x3—i4 x1=ii2 ;=i~3 ~1985 2.02 050 402* 077 045 1871986 --- 082 -- -- - -

1987 246 0.58 3Q4* 221 074 2.401988 0.71 119 0.58 002 033 0.311989 204* 138 351* 023 0.21 044All years 309 160 406* 163 049 1.75

Subscripts to Xs and null hypotheses indicate the following.— intervention only by Federal Reserve= intervention only by Bank of Japan= intervention by both the Federal Reserve and Bank of Japan= no intervention by either central bank

* statistically significant at the 5 percent level

tral bank intervened alone, days when both in- change rate in only three cases: in 198t andtervened, and days when neither intervened. 1988, when the Federal Reserve and Bundes-I ables 1 and 2 report means of the daily abso- bank both intervened on the same da and inlute changes in the PM/S and yen/s exchange 1985 when the Bundesbank acted alone, therates, respectively, and tests for equality of two cases of significant effects when both cen-means between tarious categories. ‘tables 3 and tral banks intervene is consistent with the no-4 iepeat these categories for data on the van- tion that joint actions are associated withances of daily exchange rate movements. In all significantly larger exchange rate movementcases, the data are based on absolute values of because they give signals about the futuredaily changes because we do not know the course of monetary policy and its likely effectsspecific intentions of intervention and our in- on the exchange rate. The e results also areterest merely is in a central banks ability to consistent with the finding that coordinated in-change the exchange rate. ter~entionshave larger effects on the exchange

rate than unilateral interventions.20In table 1, the data show that intervention

was associated with a significant effect on the In table 2, the effects of intervention on themagnitude of the daily change in the P 1/5 cx- yen/S exchange rate are how n to be much

20See, for example, Dominguez (1990) or Loopesko (1984).

Table 3Variances of Absolute Changes in DM/$ Exchange Rate

‘2

Year “1 “2

1985 0.00018 0.00028 0.00220 0.000721986 0.00022 0.000301987 0.00021 0.00016 0.00028 0.000121988 0.00021 0.00013 0,00032 0.000091989 0.00011 0.00012 0.00018 0.00014All years 0.00015 0.00019 0.00050 0 00030

F-statistics for Equality of Variances2 ‘2 ‘2 ‘2 2 2 2 ‘2 ‘2 2 2 ‘2

Year S1=S4 82=84 S3=S4 S1=S2 S2=S3 81=83

1985 4.11 2,56’ 3.04’ 1.60 760’ 12.50’1986 --- 1.371987 175 1.31 234’ 1.33 1.78 1,341968 2.46 1,53’ 37r 1.61 242~ 1.501989 1.35 1.23 1.28 110 1 57 1 72Aliyears 1.97 16r 1.66 122 269 3.27

Subscripts to ~‘s and null hypotheses indicate the following-~ intervention only by Federal Reserve

intervention only by Bundesbank

= intervention by both the Federal Reserve and Bundesbank= no intervention by either central bank

- = statistically signiticant at the 5 percent level

stronger: in every year but 1988 (when inter- the largest significant average absolute dailyvention generally was limited), intervention change in the exchange rate is found in 1988:either by the Fed alone or in concert with the 0.0143 pfennig. Based on an average value ofBank of Japan on the same day apparently was 1.76 for the PM/S exchange rate in 1988, anassociated with significantly larger changes in average absolute change of 0.0143 pfennig sug-

the exchange rate.21 It should be noted, gests that joint intervention is associated withhowever, that the significance of the Fed’s average absolute changes in the exchange rateunilateral interventions is based on only 26 ob- that are 0.81 percent greater than the changes

servations during the entire 1985-89 sample that occur on days without intervention, Changesperiod. Indeed, these tables show that the Fed of similar magnitudes are found for the yen/Sintervened infrequently during the period, rate. Overall, the economic significance of these

results appears to be small despite the occasion-Having found statistical significance for the ef- . . .

ally high level of statistical significance.fectiveness of intervention activities during cer-tain periods, the potential economic significance With respect to variances of daily changes inof these effects is the next question to be inves- the exchange rate, reported in tables 3 and 4, atigated. Looking, for example, at the effects of similar story carries through. For the nM/Sjoint intervention by the Fed and Bundesbank, rate, most of the significant differences between

21 For more on this story and related evidence, seeCbstfeld (1991) and Dominguez (1990).

47

Table 4Variances of Absolute Changes in Yen/S Exchange Rate

‘2 ‘2 ‘2 ‘2

Year S2 53 541985 7.96 2.01 6.32 1 371986 --- 1.41 --- 1.571987 246 069 1.55 0.741988 0.75 168 0.75 0591989 1 58 1.47 213 0.51All years 3.87 1 34 216 1.02

F-statistics for Equality of Variances2 2 ‘2 ‘2 ‘2 ‘2 ‘2 “2 “2 ‘2 2 ‘2

Year ~1=~4 S2=S4 53=54 ~1=~2 ~2=~3 ~1=~31965 5.80’ 1.46 460’ 397 3.15’ 1261986 .-- 1 11 --- --- --- -

1987 335 1.07 2.11’ 357’ 225’ 1591988 1 28 2.87’ I 28 2.25 2.24 1.001989 3.0~ 2.87’ 416’ 1 08 1.45 1.35

All years 380’ i.ar 2.t2 2.88’ 161’ 179

Subscripts to ~‘s and null nypotheses indicate the ‘ollowing

5’ = intorvont!on only by Federal Reserve

= intervention only by Japanintervention by both tie Fedoral Reserve and Japan

-- no intervent.on by either centra~bank= stattstically significant at tile 5 percent eve’

days of no intervention and some interventionoccur when the Federal Reserve and the Bun-desbank both act on the same day. For theyen/S rate, however, significant differences arefound for either the Fed or Bank of Japan act-ing unilaterally as well as for their joint actions.Intervention in all cases, however, seems to heassociated with higher, not lower, variance ofdaily exchange rate changes. Again, cautionshould he exercised before attributing cause-and-effect to these findings: an equally plausi-ble, hut untestable, hypothesis implies that vola-tility might have been even greater on thesedays had it not been for the intervention.

CONCL USI.QNS

The world’s major central banks occasionallyintervene in foreign exchange markets to affectthe value of one currency relative to another.Since 1985, the official reasons for these activi-ties changed from reducing the dollar’s value to

stabilizing it within some unspecified range ofvalues. Because much of the related data areconfidential and the stated objectives of inter-vention often are vague, researchers have beenlimited in their ability to answer a fundamentalquestion: Poes intervention work?

Using data recently released by the FederalReserve, the answer seems to be that interven-tion is associated with significantly larger dailychanges in the exchange i-ate when the FederalReserve and a foreign central bank both intei-vene on the same day. This conclusion holdseven though changes in the relative money sup-plies of two countries were described as theprimary factor behind a change in the exchangerate and the Federal Reserve routinely sterilizesits intervention, Consistent with other work oncoordinated intervention, it appears as if centralbank actions are enhanced by the announce-merit of joint actions that send a stronger signalto the market about the future course of mone-tary policy and its possible effects on the ex-change iate.

Almekinders, Geert J,, and Sylvester C.W. Eijffinger. “Empir-ical Evidence on Foreign Exchange Market Intervention:Where Do We Stand?” We/twirtschaftliches Archly (Heft 4,1991), pp. 645-77.

Batbach, Anatol B. “The Mechanics of Intervention in Ex-change Markets,” this Review (February 1978), pp. 2-7.

Batten, Dallas S., and Mack Ott. “What Can Central BanksDo About the Value of the Dollar?” this Review (May1984), pp. 16-26.

Batten, Dallas S., and Michael 1. Belongia. “The RecentDecline in Agricultural Exports: Is the Exchange Rate theCulprit?” this Review (October 1984), pp. 5-14.

Belongia, Michael T. “Prospects for International PolicyCoordination: Some Lessons from the EMS,” this Review(July/August 1988), pp. 19-29.

Bernanke, Ben, and Frederic Mishkin. “Guideposts and Sig-nals in the Conduct of Monetary Policy: Lessons from SixIndustrialized Countries,” unpublished manuscript (1991).

Bordo, Michael D.. and Anna J. Schwartz, “What Has For-eign Exchange Market Intervention Since the Plaza Agree-ment Accomplished?” NBER Working Paper No. 3562(December 1990).

Dominguez, Katherine Mary. “The informational Role of Off i-cial Foreign Exchange Intervention: An Empirical Investiga-tion,” Working Paper, Harvard University (November1988).

_______ “Market Responses to Coordinated Central BankIntervention,” Carnegie-Rochester Conference Series onPub/ic Policy, Vol. 32 (Spring 1990), pp. 121-64.

Federal Reserve Bank of New York. “Summary of Results ofU.S. Foreign Exchange Market Survey Conducted in April1989,” (September 13, 1989).

Funabashi, Yoichi. Managing the Del/ar: From the Plaza tothe Louvre (Institute for International Economics, 1988).

The discussion in the text takes the conven-tional view that intervention is sterilized if thedomestic central hank’s activities have no net ef-fect on the domestic money supply. As this ap-pendix shows, however, actions by the FederalReserve to leave the U.S. money supply un-affected by intervention still have the potentialto affect a foreign money supply and both for-eign and U.S. interest rates. In one case, the ra-tio of the foreign to U.S. money supplies will be

1Iam indebted to Manfred J.M. Neumann for making thisobservation and to Anatol B. Balbach and R. Aiton Gilbertfor developing these points in greater detail.

Henderson, Dale W. “Exchange Market Intervention Opera-tions: Their Role in Financial Policy and Their Effects,” inJohn F.O. Bilson and Richard C. Marston, eds., ExchangeRate Theory and Practice, (University of Chicago Press,1984).

Krueger, Anne 0. Exchange-Rate Determination (Cambridge:Cambridge University Press, 1983).

Loopesko, Bonnie E. “Relationships Among ExchangeRates, Intervention and Interest Rates: An Empirical Inves-tigation,” Journal of /nternational Money and Finance (De-cember 1984), pp. 257-77,

Meltzer, Allan H. “Some Empirical Findings on DifferencesBetween EMS and Non-EMS Regimes: Implications ForCurrency Blocs,” Cato Journal (FaIl 1990), pp. 455-83.

Mussa, Michael. “Empirical Regularities in the Behavior ofExchange Rates and Theories of the Foreign ExchangeMarket,” Carnegie-Rochester Conference Series on PublicPolicy, Vol. 11(1979), pp. 9-57.

Neumann, Manfred J.M. “Intervention in the Mark/DollarMarket: the Authorities’ Reaction Function,” Journal of/n-ternational Money and Finance (August 1984), pp. 223-39.

Obstfeld, Maurice. “The Effectiveness of Foreign-ExchangeIntervention: Recent Experience 1985-88.” NBER WorkingReprint No. 1525 (February 1991).

Rolnick, Arthur J., and Warren E. Weber. “A Case for FixingExchange Rates,” 1989 Annual Report, Federal ReserveBank of Minneapolis.

Ungerer, Horst, Owen Evans, Thomas Mayer and PhilipYoung. The European Monetary System: Recent Develop-ments, Occasional Paper No. 48 (International MonetaryFund, December 1986).

Weber, Warren E. “Do Sterilized Interventions Affect Ex-change Rates?” Federal Reserve Bank of MinneapolisQuarterly Review (Summer 1986), pp. 14-23.

Williams, Jeffrey. The Economic Function of Futures Markets

(New York: Cambridge University Press, 1986).

altered, thereby opening a channel throughwhich sterilized intervention, as conventionallydefined, can affect the exchange rate. In theother case, domestic monetary policies that pega short-run interest rate will require less-than-complete sterilization to keep interest rates un-changed.

Consider first figure Al, which extends theanalysis of figure 2, panel B, in the text. New

Figure AlSterilized Foreign Exchange Intervention by the Federal Reserve with Effects onGerman Money Supply: Fed Holds DM Deposits at the Bundesbank

Federal Reserve U.S Commerciat German CommercialBanks (FRB) Banks (cb) Bundesbank (B) Banks (Gob)

Assets Liabilities Assets Liabilities Assets Liabilities Assets Liabilities

DM deposits + Reserves + Reserves + DM deposits — Reserves — DM depositsatB ofcb ofFRB ofUS.

correspondentDM deposits Reserves bankat Gob of Gcb

— U.S Treasury Reserves — Reservessecunties of cb

+ U & Treasuryseounfies

entries to this original figure are shown in ital-ics. Recall that the point of this figure originallywas to show how unsterilized interventionwould affect the exchange rate by expandingthe U.S. money supply relative to the Germanmoney supply.

Picking up the story at this point, the FederalReserve could sterilize its intervention by sellingU.S. Treasury securities from its portfolio toU.S. commercial banks. This would lead to allfour italicized entries in the balance sheets ofthe Federal Reserve and U.S. commercial banks:the Fed’s balance sheet would shrink withdeclines both in its Treasury security assets andits reserves liabilities, while commercial bankswould substitute ‘treasury securities forieserves in their portfolios. On net, the sale ofU.S. Treasury securities to commercial banksand the consequent reduction in reserves willoffset the increase in reserves associated withthe initial intervention action and leave the U.S.money supply unchanged. ‘I’hus, in the conven-tional sense, the Fed has sterilized its interven-tion and, again in the conventional view, theexchange rate should be unaffected.

The figure shows, however, that if the Fedmerely holds its acquired DM as a deposit atthe Bundesbank, the Fed’s activities will havereduced the German money supply. This is indi-cated by the declines in the reserve liabilities ofthe Bundesbank and reduction in the depositsof German commercial banks. By reducing theGerman money supply relative to the U.S.

money supply, the Fed’s actions should lead toan increased nM/s exchange rate even thoughthe intervention was “sterilized” in the UnitedStates.

Holding DM deposits at the Bundesbank,however, would be somewhat unusual for theFed; typically, it invests its non-interest-earningriM deposits in interest-earning DM-denominatedsecurities. In this more usual case, the Fed’s“sterilized” intervention will not affect the Ger-man money supply if it buys German bonds inthe open market. If it were to buy them direct-ly from the Bundeshank, however, the Germanmoney supply still would be reduced as in theearlier example.

To see the mechanics of this effect, considerfigure A2. This is a reproduction of figure Al,supplemented by additional transactions shownin italics, Picking up at the point where figureAl ended—the Fed has sterilized its interventionand is holding a 1DM deposit at the Bundesbank—the Fed now buys German bonds in the openmarket from German commercial banks. Whenthe Fed’s transaction is complete, its 1DMdeposits at the Bundeshank fall, the bond hold-ings of German commercial banks fall and thereserves of the German banking system rise; onnet, these machinations offset the decline in theGerman money supply that would occur if theFed merely held its riM deposits at the Bundes-bank, In the United States, the Fed’s actionssubstitute interest-earning German bonds in itsportfolio in place of the non-interest-earning riM

Figure A2

Federal ReserveBanks (FRB)

U.S. CommercialBanks (cb)

Assets Liabilities Assets Liabilities

+ DM deposits + ReservesatB ofcb

—U.S. Treasury —Reservessecurities of cb

+ Reserves

— DM depositsat Gcb

— Reserves

+ OM depositsof FRB

— Reservesof Gob

—DM depositsof FAa

—Reserves —GM depositsof U.S.correspondentbank— German

bonds

+Reserves

deposits and leave the U.S. money stock un-affected. Thus, in the case in which the Fedeventually holds German bonds purchased inthe open market, its intervention is sterilized inthe sense of causing no change in the ratio ofGerman to U.S. money supplies.

Before considering the last case, in which theFed buys German bonds directly from the Bun-desbank, we need to ask whether’ this interven-tion really is sterilized in the sense that itshould produce no effect on the exchange rate.Although relative money supplies are unaffect-ed, the relative bond holdings of the centralbanks and the public have been altered andthese portfolio substitutions can affect U.S. andGerman interest rates. In the U.S. market, forexample, the Fed has induced U.S. banks to in-crease their holdings of U.S. Treasury securities;presumably, it did so by bidding ‘treasuryprices down, thus raising U.S. interest rates.Conversely, it bought German bonds, pr’esuma-

bly raising their prices and lowering German in-terest rates. Other things the same, thesechanges in interest rates should make capitalflow into the United States and raise the dollar’svalue against the riM. Thus, an interventionthat was designed to reduce the dollar’s valuetends to raise it because of these interest rateeffects—even if the U.S. and German monetarybases are unchanged.2 In addition to showingthat it is possible for <‘sterilized” intervention toaffect the exchange rate, this example alsoshows the effect to be in the opposite direction

from the intent of the intervention. As a practi-cal matter, these interest rate effects are likelyto be small, hut it is worth noting their exis-tence as a possible channel for exchange rateeffects from sterilized intervention.

Finally, consider the consequences of the Fedusing its CM deposits at the Bundesbank to pur-chase German bonds directly from the Bundes-bank, This case, shown in figure AS, has the

2The story actually is more complicated. First, to the extentthat the Fed’s initial (pre-sterilization) reserves injection isassociated with a “liquidity effect,” the subsequent rise inU.S. interest rates discussed at the end of this story maylust offset the earlier interest rate decline and leave U.S.interest rates, on net, at their initial levels.

It also is not possible to isolate whether intervention ofthis sort will have effects on the spot or forward exchangerate. With covered interest parity, the dollar’s forward

premium will be reduced through a rise in the dollar’s spotrate (as discussed in the example), a decline in the for-ward rate, or both. If there is no signaling effect, the bulkof the change will likely occur in the spot rate. Conversely,a strong signaling effect would likely affect the forwardrate while having little effect on the spot rate. Whichevercase prevails, the effects on the dollar’s value are unlikelyto coincide with the intent of the intervention.

Sterilized Foreign Exchange Intervention by the Federal Reserve without Effectson German Money Supply: Purchase of German Bonds in the Open Market

Bundesbank (B)

Assets Liabilities

German CommercialBanks (Gcb)

Assets Liabilities

+U.S. Treasurysecurities

+ German bonds

— DMdepositsat B

+ Reserves ofGcb

Figure A3Sterilized Foreign Exchange Intervention by the Federal Reserve with Effects onGerman Money Supply: Purchase of German Bonds from the Bundesbank

Federal Reserve U S Commercial German Commercial

Banks (FRB) Banks (cb) — Bundesbank (B) Banks (Gcb)Assets Liabilities Assets Liabilities Assets Liabilities Assets Liabilities

+ GM deposits + Reserves + Reserves German + GM deposits Reserves — GM depositsat B otcb bonds of FAB of US.

correspondent— GM deposits Reserves bank

at Gob of Gcb

— US Treasury Reserves Reserves —DM depositssecurities of ob of FRB

+US. Treasurysecurities

German bonds

— DIA deposits

at B

Bundesbank exchanging German government before, the exchange rate would be expected tobonds for its CM deposit liabilities to the Fed. change because the ratio of U.S. to GermanThis exchange, howexer, leaves in place the money upplies will be affected The upshot ofdecline in deposit and reserves at German com- these three cases is that what is typically calledmercial banks originally shown in figure Al “sterilized” intervention has a variety of chan-and, hence, the German money stock also nels through which the exchange rate could bedecline as it did in that first example. And, as affected.