Economic Review: Fall 1984, Number 4 - FRASER · 2016-01-06 · Opinions expressed in the Economic Review do not necessarily reflect the views of the ... have raised domestic real

Opinions expressed in the Economic Review do not necessarily reflect the views of themanagement of the Federal Reserve Bank of San Francisco, or of the Board of Governors ofthe Federal Reserve System.

The Federal Reserve Bank of San Francisco's Economic Review is published quarterly by the Bank'sResearch and Public Infonnatibn Department under the supervision of Joseph Bisignano, Senior VicePresident and Director of Research. The publication is edited by Gregory J. Tong, with the assistance ofKaren Rusk (editorial) and William Rosenthal (graphics).

For free copies of this and other Federal Reserve publications, write or phone the Public InformationDepartment, Federal Reserve Bank of San Francisco, P. O. Box 7702, San Francisco, California 94120.Phone (415) 974-3234.

2

I. Budget Deficits, Exchange Rates and the Current Account: Theoryand U. S. Evidence ................................................................................5

Michael Hutchison and Charles Pigott

II. The Real Interest Rate/Budget Deficit Link:International Evidence, 1973-82 ................................................... 26

Michael Hutchison and David H. Pyle

III. Money Supply Announcements, Forward Interest Rates andBudget Deficits ...................................................................................36

John P. Judd

Editorial Committee:Michael Keeley, Bharat Trehan, Michael Keran and Adrian Throop.

3

Michael Hutchison and Charles Pigott*

Recent U.S. current account deficits represent a net inflow offoreignsaving to help finance our government budget deficits. Budget deficitshave raised domestic real interest rates and the real value ofthe dollar,causing the deterioration in our current account balance. Eventually ourreal interest rates shouldfall back toward world levels asforeign savinginflows increase. But, contrary to conventional wisdom, large U.S.current account deficits will probably remain as long as our budgetdeficits persist.

Over the last several years, the United Stateseconomy has seen record highs in federal budgetdeficits, real interest rates, the dollar and tradedeficits. Last year, the federal budget deficit climbedto $195 billion, more than 5 percent of GNP-thehighest rate of the postwar era. The federal deficitthis year is expected to be about $170 billion. U.S.real interest rates remain far above past historicalaverages, while the dollar recently reached a ten-yearhigh against the German mark and British pound.Our current account deficit, measuring the differencebetween our imports and exports of goods and ser­vices, will probably reach $90 billion for 1984­an all-time record. And to finance this deficit, ournation will have to borrow an unprecedented amountfrom abroad.

Many solutions have been proposed to deal withindividual problems arising from these conditions.Growing fears that trade deficits, and the associatedloss of jobs and markets for U.S. export industries,are' 'de-industrializing" America have added con-

* Economist and Senior Economist, respectively.We would like to thank the Editorial Committeeand Hang-Sheng Cheng for helpful comments,and Julia Lowell for her research assistance.

5

siderably to protectionist pressures. Improving ourtrade deficit was also an objective of the recentU.S.-Japanese agreement to reduce Japan's barriersto international capital flows.

Increasingly, though, deficits, high interests rates,and the dollar are being viewed not as isolatedproblems to be dealt with separately, but as closelyrelated consequences of a common cause-govern­ment fiscal policy. Many analysts believe that fed­eral budget deficits, by pushing up domestic interestrates, are ultimately responsible for the strong dollar,itself the major cause of the nation's declining inter­national competitiveness and rising trade deficit.This view suggests that only by balancing the budgetcan the other problems be fundamentally and per­manently resolved.

Still, there remains considerable controversy overwhether the budget deficits' impacts are really thispervasive, and over what continued deficits maymean for the future. Some analysts believe thatincreased business investment yields resulting fromrecent tax cuts, rather than budget deficits them­selves, are mainly responsible for our high realinterest rates. Others argue that the high dollar ismore a reflection of foreigners' flight from politicaland economic problems in their countries than of

the attractiveness of high U.S. interest rates. Thereis also a very widespread view that our. currentaccount deficits cannot be sustained much longer,leading to predictions that the dollar must soon fallsubstantially to bring our international trade back tobalance. Yet others argue that our external deficits,far from being an economic problem, have actuallybeen beneficial by allowing domestic investment toremain strong despite the increasingly enormouscredit demands from the government.

These controversies raise the fundamental ques­tion that is the focus of this paper-what are theimpacts, short-term and long-term, of budget defi­cits in an open economy, one with trade and financialties to abroad? In attempting to answer this ques­tion, our analysis begins with the observation that thecurrent account deficit is not only the differencebetween our exports and imports, it is also thechannel through which foreign saving is broughtinto our economy to help meet the credit demands ofboth the government and the private sector. In thisway, a rise in the budget deficit may easily (but notinevitably) lead to a current account deficit, depen­ding on the extent to which the government's creditdemands are met, directly or indirectly, from foreignas well as domestic sources. 1

In Section I, we develop a conceptual frameworkrelating budget deficits, interest rates, exchange

rates, and the currentaccount for an open economyunder flexible exchange rates. This framework sug­gests that budget deficits are likely initially to raisedomestic real interest rates which, in tum, push upthe real exchange rate. As time passes, this increasein the real exchange rate leads to a current accountdeficit, allowing foreign saving to supplement do­mestic saving in financing the budget deficit. Inapplying this framework to the U.S. (Section II),we argue that this sequence fits our experience ofthe last several years fairly well, suggesting thatbudget policy is indeed mainly responsible for ourcurrent account deficits.

Since ongoing budget deficits imply an ongoinggovernment need for private saving, our frameworkimplies that, in principle, they can lead to ongoingcurrent account deficits as well. For the U. S., thissuggests that our current account deficits may indeedbe sustained as long as our budget deficits remain.Furthermore, these external deficits may help re­duce, although certainly not eliminate, the economiccosts typically viewed as the consequence of budgetdeficits. In particular, the inflow of foreign savinginto our economy should allow our interest rates,and the real value of the dollar, to decline somewhat,and domestic investment to escape substantially, ifnot completely, being "crowded out" by the gov­ernment's credit demands.

I. Conceptual FrameworkIn this section, we develop a conceptual frame­

work to describe how budget deficits may influencethe current account and the channels through whichthis influence is transmitted under a floatingexchange rate regime. Our theory applies the modemasset market approach to exchange rate determina­tion (for example, Dornbusch, 1976, Isard, 1980) tothe static short-term fiscal analysis of Mundell (1962)and Flemming (1962). Our framework integratesand extends recent work on short-term dynamicadjustment of the open economy to fiscal deficits(for example, Blanchard and Dornbusch, 1984,Hodrick 1980, Sachs and Wyplosz 1984) with ananalysis of the deficits' long-term impacts. The nextsection then applies this framework to recent U. S.history. Our analysis is deliberately heuristic andfairly non-technical. More formal and technicalanalysis is relegated to footnotes and cited references.

6

A. The Accounting RelationThe basic reason that budget and current account

deficits are related is because budget deficits repre­sent a use of saving and current account deficits asource of saving. This may be seen from the nationalsaving identity:2

(I) (G-T)

(Budget Deficit)

(S - I) +(Private DomesticSaving Surplus)

(M+R-X) +(Current AccountDeficit)

The government budget deficit (expenditures lesstaxes, G T) must equal, or be financed by, theexcess of private domestic saving (S) over private

investment (I) plus the current account deficit. Thecurrent account deficit is the difference between anation's expenditures on foreign goods and services(imports, M) and net transfers (R) and foreign ex­penditures on our products (exports, X). This dif­ference must be financed by borrowing an equalamount of funds from abroad, and in this sense acurrent account deficit represents a net tlow of for­eign funds (foreign saving) to our economy.

In tlow of funds terminology, the budget deficitand private investment constitute competing "uses"of savings. The' 'sources" of this saving are privatedomestic saving (S) and the funds from the foreignsector represented by the current account deficit.

It is not only true that a current account deficitrequires a net intlow of foreign funds to finance it: anation can sustain a net financial inflow from abroadonly by incurring an equal current account deficit.Net borrowing from abroad effectively amounts toselling foreigners more "IODs" than we purchasefrom them. Overall, a nation's accounts withabroad-trade and financial-must balance. Acountry cannot be a net borrower of foreign funds(net "exporter" of IODs) without being a net im­porter of commodities and services. Thus, a nationcan draw on foreign savings for its domestic needsonly by incurring a current account deficit.

Taken by itself, an increase in government creditneeds might be met partly by borrowing fromabroad. This would seem to suggest that budgetdeficits would inevitably lead to current accountdeficits. However, the policies or other factors lead­ing to a budget deficit will often affect domesticsavings and investment as well. During a recession,for example, although the budget deficit tends torise, the private domestic saving surplus typicallyincreases even more (because of depressed invest­ment demand). As a result, the budget and currentaccount deficits generally move in opposite direc­tions over the business cycle. Furthermore, govern­ment policies underlying a fiscal deficit could havean independent impact on private saving and invest­ment that would make it unnecessary to borrowfrom abroad. Thus, while there is an importantrelation between budget and external accounts, thereis no rigid mechanical linkage between the two.This means that a budget deficit's impact on domes­tic saving and investment demand must be assessed

7

before its implications for the current account canbe determined.

B. Short-run linkagesThe purpose of this section is to trace out the

likely short-term effects of a rise in the governmentbudget deficit on interest rates, exchange rates, thedomestic saving surplus and the current account.'To this end, we begin by sketching a simple outlineof the dynamic process. We fill in the details andmodify the story in the following section.

Consider the case where the government adoptspolicies that raise the budget deficit but do notdirectly affect private sector saving, investment orthe current account. Initially, a rise in the budgetdeficit is likely to have expansionary effects ondomestic output and employment. An expandingeconomy, in turn, generates an increased privatedomestic savings surplus which, to a large extent,may absorb the additional government demand forcredit without putting significant upward pressureon real interest rates. As the standard textbookanalysis suggests, interest rate pressures are morelikely to be averted the larger the degree of unem­ployed resources in the economy and the strongerthe stimulative impact the budget deficits have onoutput.

Once the economy approaches full employmentand the initial output effects of a rise in the budgetdeficit subsides, the remaining deficit must be fi­nanced from a combination of a rise in the privatedomestic saving surplus relative to GNP and froman inflow offoreign saving. Further increases in theprivate domestic saving surplus are unlikely to beforthcoming, however, without a rise in real interestrates.4 In this instance, credit market pressures aris­ing from the tender of government securities exertdownward pressure on bond prices, and real interestrates rise as a consequence. Higher real interestrates, in turn, tend to stimulate private savings(lower interest-sensitive consumption expenditures)and slow investment outlays. Through this adjust­ment process, budget deficits may at first be financedlargely from domestic sources.

This is illustrated in Figure I, which shows thesources and uses of loanable funds in the domesticeconomy. The uses of funds represent governmentand private domestic demands for credit. The sources

Figure 1

Sources and Usesof Loanable Funds:EffectsofaRise in Government Credit Demands

and Sluggish Foreign Capital InflowsReal Interest

Rate,r

r,r,

ro

U'

Fo F, F,

Flow of Funds:Sources of Funds (S)Uses of Funds (U)

of funds (S) represent domestic private saving plusforeign saving flowing into the economy for a givencurrent account deficit. That is, S represents ashort-run sources-of-funds schedule. The economyis initially at equilibrium point a, with short-termreal interest rate r l and the uses and sources of fundsequal at Fl. The increase in the budget deficit shiftsout UO to U I as the government's demand for fundsrises. At unchanged short-term real interest rates,the total demand for funds, represented by point a' ,exceeds the available supply. Excess demand forfunds under normal conditions will increase the realinterest rate from ro to r l , moving the economyupward along SO as private savings increase, andaway from point a' to point b as private creditdemands are scaled back.

In an open economy, however, increased foreignsavings are also likely to partly finance increaseddomestic budget deficits. Higher domestic real in­terest rates, ceterius paribus, will cause investors toattempt to shift out of foreign assets and into domes­tic assets in order to take advantage of higher do­mestic real yields. The rise in demand for domesticassets, in tum, will put upward pressure on thedomestic currency in the foreign exchange market.

8

As investors move to sell foreign currency for do­mesticcurrency and use the receipts to purchasehigher yielding domestic bonds, they bid up theexchange rate.

Real domestic currency. appreciation associatedwith higher real interest rates also represents a risein the price of domestically produced goods relativeto those produced abroad. This weakens exportdemand and spurs imports, causing the current ac­count balance to deteriorate gradually. Current ac­count deterioration, in tum, is the mechanism thatallows foreign savings to begin to supplement do­mestic savings in financing domestic governmentbudget deficits and private domestic investment. Inour diagram, this shifts the short-run sources offunds schedule from SO to SI and further to S2. Theinflow of foreign savings, which gradually developsas the current account declines, represents an in­crease in the net supply of domestic assets held byforeign investors.

Thus, although increased foreign demand fordomestic assets may at first result primarily in priceeffects (exchange rate appreciation), the supply (netstock) of domestic securities available to foreignerswill also begin to increase as domestic currencyappreciation in real terms causes the current accountto decline and foreign capital to flow into the do­mestic economy. However, as the sources of fundsschedule shifts outward following the inflow offoreign savings, domestic credit market pressuresshould ease, allowing real interest rates to fall backtoward the world level. This is shown in the diagramas the movement from rl to r2 , and the economymoves through a succession of new short-term assetmarket equilibrium positions represented by pointsb, c and d. This suggests that the inflow of foreignsavings will play an increasingly important role infinancing domestic budget deficits in an open econ­omy.

The preceding describes the general pattern ofinitial adjustment in an open economy following arise in the government budget deficit. However, thesame pattern would also result from a business taxreduction or other policy that increases the after-taxreturn on domestic investment but does not neces­sarily increase the budget deficit. In those cases,private domestic investment demand would rise,leading to real interest and real exchange rate in­creases similar to those just described. Moreover,

when a policy increases both the after-tax return ondomestic investment and the budget deficit, an ad­ditional reinforcing and magnifying effect on interestrates and exchange rates would likely be introduced.We now consider in more detail two features of thisprocess: the different roles played by long-term andshort-term interest rates and the role played by lagsin the adjustment of the current account.

Exchange Rates and Interest Rates. The extent towhich the real exchange rate initially appreciateswill depend upon market expectations of the durationof government budget deficits and how they in­fluence short-term and long-term real interest rates.The relationship between the real exchange rate andreal interest rates may be seen explicitly by con­sidering the equilibrium condition for internationaltrade in assets:

r* = r - (q - qC) (I )

where

r, r* = log of one plus the domestic and foreignreal interest rates (yield to maturity), re­spectively.6

q, qC = log of the spot and expected future realexchange rate, respectively. (The real ex­change rate is defined as the nominal ex­change rate-foreign currency per unit ofdomestic currency-deflated by the ratio ofthe foreign to domestic price levels.) Thebonds underlying r, r* are of equal maturityand also correspond to the time horizon ofthe real exchange rate expected in the future.

The left-hand side of Equation (I) represents theexpected real return (risk adjusted)5 available toforeign investors for holding a foreign bond, r*.The return available to foreign investors for holdingdomestic bonds has two components: the yield onthe domestic bond, r (denominated in domestic cur­rency), less the expected future depreciation of thedomestic exchange rate. The expected percentagereal depreciation of the currency, in tum, equals thedifference between the currently observed real spotexchange rate (in log form) and the spot rate expectedto prevail at the point the domestic bond maturesand the foreign investor converts the proceeds fromdomestic to foreign currency.

Equalized expected real returns for similar bondsacross countries is the condition for international

9

capital market equilibrium. When Equation (I)

holds, this condition is met. The expected real returnavailable on foreign bonds will then equal the ex­pected real return on domestic bonds, adjusted forthe expected change in the purchasing power ofthecurrency. Investor arbitrage in international capitalmarkets will cause this equilibrium condition tohold almost continuously.

This equilibrium condition should hold for thefull term structure of real interest rates. For example,in equilibrium, a 10 percent rate of return· on aone-year foreign security and a 12 percent rate ofreturn on a one-year domestic security indicates thata 2 percent depreciation of the domestic currency isexpected by investors over the course of the year.On the other hand, 10 percent and 12 percent an­nualized real yields onfive-year foreign and domes­tic securities, respectively, suggest that an average2 percent rate of currency depreciation per year isexpected by investors over a five-year period, indi­cating a total expected depreciation to maturity ofapproximately 10 percent.

That is important because it suggests that budgetdeficit policies that lead to an increase in long-terminterest rates are likely to have significantly largerimpacts on real exchange rates than a policy givingan equal rise in short-term interest rates. For exam­ple, consider a one percentage point rise in thedomestic 5-year real interest rate, with no change inthe foreign real interest rate and no change in thereal exchange rate expected to prevail five years inthe future (qe). This would cause investors to bid upthe real value of the domestic currency (q) by 5percentage points. The real value of the spot ex­change rate in this case rises to that point above theexpected future value of the exchange rate wherethe expected depreciation of the domestic currency(five percent over a five-year period) just offsets theadditional return on the domestic security. In com­parison, one percentage point rise in the domesticone-year real interest rate (with no change in otherexpected future short-term interest rates) would leadto a one percent appreciation of the domestic ex­change rate, ceterius paribus, thereby setting up anexpected depreciation of one percent over the yearand restoring net yields on foreign and domesticsecurities to equality.7

Budget deficit policies that are not expected to be

reversed in the foreseeable future and that lead tosignificant increases in long-term real rates of inter­est would, therefore, probably result in a muchgreater appreciation of the domestic currency thandeficits that are expected to be temporary and in­fluence mainly short-term rates. Hence, market ex­pectations of the duration of budget deficit policiesin the economy and their influence on the termstructure of interest rates will playa major role indetermining the extent to which the domestic cur­rency appreciates.8

Current Account Adjustment Lags. The path ofthe economy we have sketched is crucially depen­dent upon sluggish current account adjustment. Inparticular, lags in the adjustment of the currentaccount to a rise in the budget deficit are primarilyresponsible for the rise, or "overshooting", of do­mestic real interest rates and the real exchange rateabove their long-term values. That is, given suffi­cient time, budget deficits may raise the currentaccount deficit, either directly (as fiscal policiesdirectly alter export supplies and import demands),or indirectly through their impact on real interestrates and the real exchange rate. Typically, though,these adjustments in exports and imports occur onlyafter a considerable lag. In the interim, budget defi­cits must be financed primarily from private domes­tic surplus saving; domestic real interest rates thenmust rise to generate this surplus, driving the realexchange rate above its long-term value.

This process is usually rather lengthy. For exam­ple, exports and imports generally take two or moreyears to respond fully to changes in real exchangerates. 9 However, as the current account graduallyadjusts and foreign savings do begin to supplementdomestic sources in financing the budget deficit,pressures on domestic real interest rates are apt toease. Real interest rates and the real exchange rateare then likely to begin falling back toward theirlong-term values, a process only completed whenthe current account has fully adjusted. (See Box I.)Conversely, "overshooting" of interest rates andthe exchange rate would not occur if the currentaccount were to adjust immediately. Such immediateadjustment implies that foreign savings could beinstantaneously drawn upon to finance the rise inthe budget deficit.

10

C. Long-Term ConsequencesOur analysis suggests that increasing inflows

of foreign funds through the current account willultimately ease pressures on domestic real interestrates and the real exchange rate. Where will thisprocess end? And what are the long-term economicconsequences of ongoing budget deficits? Thesequestions raise several issues: the sustainabilityof current account deficits, the long-term conse­quences of ongoing budget deficits for domesticinvestment, future output and the economic well­being of the nation's residents, as well as the ulti­mate level ofdomestic real interest rates and the realexchange rate.

In considering these issues, we assume that thegovernment has instituted policies that lead to apermanent budget deficit fixed at some constantfraction of GNP. We also presume that domesticprivate saving does not rise enough to finance thedeficit fully (so there is, at least potentially, a per­manent need for foreign saving inflows). Full em­ployment is also assumed since we are consideringlong-term consequences.

Sustainable? There is a widespread convictionthat a nation's current account cannot sustain adeficit on an ongoing basis, and ultimately mustcome back into balance. This view implies that anongoing budget deficit would eventually have to befinanced entirely from the surplus saving (S-I) ofthe domestic private sector; domestic investment(or consumption) ultimately would have to fall tofinance the budget deficit.

This presumption would certainly be valid in aworld in which there was no saving or growth.Foreign wealth would then be constant, yet eachyear foreigners would have to allocate an additionalportion of that wealth to finance another nation'scurrent account deficit. Since foreigners wouldeventually run out of funds to lend, an on-goingcurrent account deficit-indeed, an on-going budgetdeficit-would be impossible in a static worldeconomy.

In a growing world economy, however, foreignsaving (which represents the increase in foreignwealth) could finance a nation's current accountdeficit indefinitely (provided it did not exceed for­eign saving). In this way, foreigners could lend to a

11

12

13

nation year-after"year while. maintaining constantthe share of their wealth devoted to that purpose(this .share would, of course, be gre:iter the largerthe current account deficit in relation to foreignsavings). iO .Thus, current •accountdeficit§are notintrinsically unsustainable in a growing worldeconomy. II

Willing? The reaLlimit to the sustainability pfacurrent account deficit is likely to be the willingnessof foreigners to lend their savings to the. nationincurring. it. Lending. to another l1ation.(itsgovem­ment or its citizens) often involves .certain risks~

known as "country risks"~that may limit the sizeof the current account foreigners are willing to .fi­nance on an on-going basis.

These country risks are of three basic types. Thefirst, known as "sovereign risk", reflects the pos­sibility that the govemment of the borrower willdefault, that is, repudiate its own and/or its citizens'foreign debts. The second, "transfer risk", refers tothe possibility that the borrower will be unable toobtain the foreign exchange needed to repay a for­eign debt (when the loan is extended in foreigncurrency). This is most likely to occur when anation uses exchange controls to maintain an over­valued exchange rate. Transfer risk has proven to bethe major country risk incurred in lending to devel­oping nations. Finally, foreign (as well as domestic)lenders may also face possible losses from certainmacroeconomic policies of the.borrower's govern­ment. The most serious of these risks is from policiesthat lead to unanticipated inflation and currencydepreciation and thereby reduce the real value of thefunds lent. 12

The degree of country risk critically affects theinterest rate a nation must pay to borrow from abroad,as well as the amount of funds it can obtain. Wherethis risk is present, a country must· compensateforeign lenders by paying them a real interest rate(adjusted for expected exchange depreciation~see

Equation 2) above that prevailing abroad. This dif­ference, the "country risk premium", is analogousto the. yield premium paid by Baa over Triple-Adomestic bonds.

Furthermore, the amount of funds a nation canborrow from al:>road on an ongoing basis (the sus­tainable current account deficit) to help finance agiven budget deficit will be smaller the greater is thecountry risk and associated risk premium. Indeed, if

the risk weregrearenough, a nation could find itselfunable to sustain any current account deficit. Anongoing budget deficit would then raise domesticreal interest rates permanently above world levels,toalevelthatreduced private investment relative toprivate/saving enough to finance the deficit entirelyfrom domestic sources. Thus, the higher the countryrisk, the more closely a budget's long-term impactsol1inrerest rates and investment will resemble thosefora closed economy, and the more domestic in­vestment isultirl1ately constrained by the availabledomestic savings (less the budget deficit).

Where there is no country risk, a budget deficit'slong-term impacts on domestic interest rates andinvestment are likely to be very different. In thatca<;e, foreigners would be willing to lend to thenation on the same terms as they receive at home.The resulting situation is analogous to that facingindividual regions of the U.S. economy. Within theU.S.,the residence of a borrower does not by itselfusuallyatfect the terms of a loan, nor does it gener­ally affect the willingness of a lender to extendcredit. Hence, an Alaskan firm can borrow on thesame terms as a similar Illinois firm, and neitherAlaskan· nor Illinois savers generally have any"habitat" preference for investments in their ownstates' firms. In effect, all borrowers in a given typeof activity regardless of their location face a singlenational interest rate. In an international context,the absence of country risk thus means that a nationwith an ongoing budget deficit will see its domesticreal· interest rates ultimately fall back to worldlevels. 13

Furthermore, absence of country risk also impliesthat the level of domestic investment will be deter­mined by its profitability relative to investmentsabroad, not by the level of private domestic savingless the budget deficit-as is true for a closed econ­omy.. The on-going current account deficit thusequals the difference between the profitable level ofdomestic investment and domestic saving less thebudget deficit, and will be financed by foreigners atw()fldreal interest rates. Again, the level of savingof Alaskan residents was not a serious constraint oninvestment in its oil fields; the oil fields were devel­oped primarily with funds from non-residents.

In sum, where country risk is small, a budget

14

deficit's long-term impact on domestic investmentwilldependmainly upon how the policies generatingthedeficitaffectthe profitability ofdomestic invest­ment. The more these policies enhance the profitsfronrthatinvestillent,thehigher the level of invest­mentinthenation,andthelarger its national outputandcurrentaccountdeficits in the long-term. Onthis basis, deficits resulting from business tax cutscould raise a nation's share •of world investment.Deficits that raise the demand for products the nationhaliacomparativeiadvantage producing may alsotend to encourage domestic investment by raisingthe prices ofthOlie products and hence the profitsavailable to those producing them. Clearly though,budget deficits may also be generated by policiesthatreduce the yield to domestic investment and leadtoafall in its level relative to that abroad. Whencountry risk is absent, therefore, the "content"rather than the size of the budget deficit determinesits long-term impact on investment (and nationaloutput) and hence plays a critical role in determiningthe size of the ongoing current account deficit.

A similar observation applies to the long-termimpact of budget deficits on the real exchange rate.The real exchange rate is simply the nominal rate, 'deflated" by the ratio of the domestic to the foreignprice level. As such, it effectively measures thevalue of thenation'li products in terms of thoseproduced abroad (that is, the relative price of a"basket" of home-produced goods in terms of a"baliket" of foreign products). Ultimately, thisvalue will be determined by commodity demandand liupply for these products. Accordingly, thedeficit's long-term impact on the real exchange ratedepends on how the measures underlying it affectthe· demand and supply for home versus foreignproducts. A deficit generated by measures that shiftdemand toward domestic products (for example byincreasing expenditures on domestically-produceddefense goods) will tend to raise their relative pricein terms of foreign products. But a deficit may alsolead to a long-term real depreciation if it shiftsdemand away from home goods, or increases theirsupply more than the demand for them. Again, thepolicies making up the deficits, rather than the defi­cits' size, are the determining factors. 14

II. Applications to the Recent U.S. Experience

Chart 1

Federal Budget Balanceand the U.S. Current Account

Billions of Dollars

The theory outlined .in the previous section leadsto specific: predictions about the wayb\.ldget deficitsare likely to affect the current account, investment~uld s<iving, and about the exchange rate ami interestrate linkages •• through which·· thispr9cess .occurs.HoW well does the United States experience, par­ticularly recently, fit t~e theory? And which as­sumpti9ns unqerlying the longer run preqictions ofthe theory seem to best fit the U.S. and its relation­ship with the rest of the world?

-50

50

90

Chart 2

The Real Dollar Exchange Rateand the U.S. Current Account

1980-82~ 100 Billions of Dollars

tively); current account deficits are expected to growto new. rec9~ds forecasts range from $80-$120billion in both years. Budget deficits and currentaccount deficits of this magnitude are unprecedented.

During the post-1973 floating exchange Tate per­iod, there has been a close correlation between thecurrent account and federal budget balances. ChartI shows the tight link between the cyclically adjustedfederal budget deficit and the current accountbal­ance of the following year (to allow for sluggishcurrent account adjustment) in the 1973-83 period.

These developments are consistent withthe pat­tern predicted by theory. In addition, the sharpdeterioration in the current account is most probablyrelated to the extraordinarily high value of the dollarin recent years. This inverse correlation is shown inChart 2. Numerous formal empirical analyses alsosuggest that the more than 40 percent appreciationof the average value of the dollar since 1980. isresponsible for the greater part of the U. S. curre.ntaccount deterioration. 15 The high level of real long­term interest rates (inflation-adjusted) prevailing inthe U.S. since 1980, in tum, may be largely respon­sible for the dollar's dramatic appreciation. This isillustrated in Chart 3, and is the conclusion reachedby a number of formal empirical studies. 16

Moreover, part of the recent pattern of U.S. in­terest rates is also consistent with the dynamic pro­cess predicted by theory. In particular, the very high

120

110

100

1983198119791977

Current Account(following year)

•

50

-50

-100

A.. The.Recent U.S. ExperienceThe recent\.lpward climb in the federal budget

deficit is in fact associated with a substantial deter­ioration in the current account of the balance ofpayments. The federal budget deficit climbed from$57.9 billion in 1981 to $1I0.6 billion in 1982, andfurther to $195.4 billion in 1983. Following a similartrend, the current account deteriorated over thisperiod from a $4.5 billion surplus in 1981, to an$11.2 billion deficit in 1982 and a $40.8 billiondeficit in 1983. In addition, while budget deficitsare expected to level off in 1984 and 1985 (theCouncil of Economic Advisors forecasts $183 billionand $180 billion deficits in 19S4 and 1985, respec-

Source: Fieleke (1984), Chart 2, p. 7. Budget data are fromSurvey of Current Business and from Commerce Depart­ment statt; current account data are from Economic Reportof the President, 1984, p. 250 (net foreign investment) andCommerce Department staff, except for 1984, which is aforecast.

80 Lo-........I-...............L....................--1I--........ -1001973 1975 1977 1979 1981 1983

Sources: Current account data-see notes to Chart 1; realdollar exchange rate data are from Morgan Guaranty TrustCompany, IM:Jrld Financial Markets.

15

25

-25

1980

Chart 4

Federal Budget Balance andLong-Term Real Interest Rates

Billions of Dollars

1979

oP""-'"

5

Percent

15

10

Sources: Federal budget balance data are from Survey ofCurrent Business; real interest rate is the difference be­tweenthe20-year constant maturity bond and the year­oveFyear percent change in the Personal ConsumptionExpenditure Index.

t;liptyinvolving.thefuturecourse of U.S . fiscal andmonetary policies and, hence, the future course ofreal interest rates. In the face of significant uncer­tai~ty,and with •• continual revisions of expectations~~<ne)\l infollllation becomes available, the exactpaths of reallong-tenn interest rates and the realexthange<iate<areconsiderably more difficult topredict than our simple theory suggests. A highdollarand high real interest rates could continue fora cO~SideffPleperiod· ~nder these circumstances.

A second factor may be the timing ofthe predict-.ed decline in real rates and the dollar. Foreign capitalinflows may not yet have reached the point wherethey can signifitantly ease pressures on U.S. capitalmarkets, particUlarly .in light of the rapid rise inprivate credit demands associated with the robustU.S. economic recovery. If this is the case, interestrates and the dollar could edge downward when therecovery matures and private credit demands abate.

In any case, the budget deficit explanation is atleastas consistent with the actual record as the mainaltern~tives that have been offered. As noted earlier,sOme have argued that the massive net inflows ofcapital mainly reflect flight into the U.S. as a "safehaven" from political and economic troubles abroadrather than a response to high U. S. real interestrates. But such a flight, while it could explain thehigh dollar and (hence) deteriorating current ac­count, would tend to lower, rather than raise, U.S.

5

o

10Percent

1983198219811980

Chart 3

Long-Term Real Interest Rates and theRea! Dollar Exchange Rate

80 ..................- .......L.--......- ........-----"-51979

Sources: Real dollar exchange rate-see notes to Chart 2;long-term real interest rate-difference between the 20­year constant maturity bond yield and the year-over-yearperCElnt change in the Personal Consumption ExpenditureIndex.

levels of short-tenn and long-tenn interest ratessince 1981 may be attributable mainly to large andincreasing federal government budget deficits (seeChart 4). The rapid runup in long-tenn rates, inparticular, is consistent with market expectations ofa long series of large future budget deficits, andassociated high future short-tenn interest rates. For­mal statistical evidence on the budget deficit/realinterest rate link is inconclusive,17 however, dueperhaps to the fact that past large budget deficitshave generally (aside from war periods) occurredduring recessions when private credit demand wasweak. In contrast to our present situation, past defi­cits typically have disappeared once the economyreached full-employment. Nevertheless, based onthe lack of a strong simple statistical correlationduring-the greater part of the post-war period, seve­ral prominent observers have contended that presentbudget deficits are not primarily responsible for ourhigh real interest rates, and therefore deny theirconsequences for the dollar and the current account. IR

In contrast to what theory suggests, there is as yetno indication that credit market conditions in theU.S. have eased with the large inflow of foreigncapital. Both real interest rates and the real value ofthe dollar continue to remain at high levels. Severalfactors may be responsible for these developments.Perhaps most important is the great deal of uncer-

1980--82=100

16

real interest rates relative to those prevailing abroad.The safe haven analysis therefore cannot be themain explanation for the events we have traced.

Another potential explanation attributes the dol­lar's strength, and highIJ.S...real interest rates (onfinancial assets), to the increased after-tax yield oninvestment in the U.S.that resulted from recentreductions in corporate taxes. As indicated earlier,such a fiscalpolicy would raise private investmentdemand and lead to essentially the same pattern ofinterest rate.,exchal1ge rate,and current accountadjustments as a. budget deficit. This explanation,however, complements rather than competes withthe budget deficit explanation. Both trace high in­terest rates and the high dollar to fiscal policy. 19

B. Long-Term ImplicationsSince there is some evidence that the theory of the

last section does apply to the present U.S. situation,it is worthwhile to consider its implications con­cerning the long-term effects of our budget deficitsshould they persist indefinitely (as most observersbelieve they will without substantial policychanges). In particular, are the large current accountdeficits the U.S. has been running really unsustain­able as many observers believe? If not, how largecould they be on an ongoing basis? And will ongoingbudget deficits inevitably mean high real interestrates, depressed investment and lower future output?

Large as U. S. current account deficits have been,they are still substantially less than the foreign sav­ing available to finance them. For example, a deficitequal to 2.5 percent of GNP-slightly less than therate projected for this year-represents about 12percent of the saving of foreign industrial nations,net of depreciation. Foreigners certainly could fi­nance U.S. deficits in this range, although the shareof their wealth they ultimately would have to devoteto claims on our country (about 12 percent) wouldcertainly be very large by historical standards. 20

Morever, any country risk associated with theU.S. is apt to be very small, indeed negligible,provided foreigners remain confident that our infla­tion will continue to be contained. Given this confi­dence, foreign willingness to lend should not be aserious constraint on the size of future U. S. currentaccount deficits.

The risk most often associated with foreign len-

17

ding-transfer risk-is apt to be negligible for theU.S. given the key international role of the dollarandthe openness ofour financial markets to foreignfinancial flows. Certainly,. sovereign risk qm alsobe neglectedJor they.S. given itslongnistory ofpolitical stability. Indeed, •. the U.S .• and the •• dollarappear increasingly to be/regarded as safe-havensfor funds from abroad. This implies that foreignerswould be willing •to .lend here • teflllsthatareatleast as favorable as those they would demand athome.

Thissceriario leaves unanticipated inflationthe major potential risk faced by foreign (and do­lllestic) lenders tathe· U.S. An unf6teseenandprolonged surge in U.S. il1flationcould serious­lyerode the purchasing power of funds lent (indollars) by foreigners. If concerns were to arise thatU.S. inflation might be rekindled, foreign reluc­tance to invest here could conceivably become aserious obstacle to the financing of our externaldeficits. However, another serious.round of inflationis only likely to occur if there is a substantial shift inmonetary policy away from its present anti-infla­tionary stance. In effect, then, country risk is apt toremain neglible for the U.S. as long as our govern­ment maintains the credibility of its anti-inflationcommitment. 21

How large? Assuming, then, thatforeigners\\iillfinance an ongoing U.S. currentaccountcteficit,how large could it be? The answer clearly dependson the size of the ongoing public sectordeficit,aswell as on how private domesticsavinganQinvest­ment are affected (See Box 2 for further details).Current projections suggest that the combined defi­cit of federal, state, and localgovernlllents willaverage 3.0-3.5 percent of GNP in coming yearsunder present policies. 22 It is also reasonable toexpect the net private saving rate to remain at itspostwar average (about 7.3 percent. of GNP) sincepast U.S. experience suggests it is both stable andnot significantly affected by budget deficits. 23

Assessing our future investment rate is more diffi­cult, since it will depend on how profit opportunitieshere cOlupare with those abroad.. Given that recentU.S. business.tax cuts have significantly raised theafter-tax return to business, the U.S. shareof worldinvestment might conservatively be projected .toremain (at least) at its past average. This would

18

19

imply (see box) a.net investIllent rate ofapproxi­mately 6.0 percent of our GNP.

Together, these (very) rough projections imply anongoing U.S. current account deficit of 1.7__2.2percent ofGNP, representing ~55-70 billion at 1.984prices and GNP. Foreigners wouldthen befipancingover half of our public sector deficit, .leavillg lessthan half to be financed from domestic sources. Ofcourse, ifdomestic saving were to rise (as supply­siders expect), the current ac<::ount deficit could besignificantly less. Alternatively, ifthe U.S.shareofworld investment were to rise, the deficit could bemuch greater. Despite these uncertainties, this ex­ercise does indicate that the persistence of budgetdeficits at current rates almost certainly will lead tounprecedentedly large ongoing U.S. current accountdeficits. Still, as we now argue, such deficits,shocking as they may seem, are not, o/themselves,necessarily harmful to our economy.

Where's the Burden? Government budget deficitsare thought to impose burdens, or economic costs,on the nation incurring them. Ofcourse, these costsmust be weighed against the benefits the policiesunderlying the deficits may bring. In this sense, thecosts of budget deficits reflect the reallocation ofsociety's resources-from future to present expen­diture and between public and private spending­rather than any misallocation of those resources, orburden to the nation as a whole (that is, present plusfuture generations). Nonetheless, conventionaltheory suggests that deficits will impose costs oncertain sectors and individuals-manifest in termsof higher real interest rates,.Iower domestic invest­ment, and lower private consumption for futuregenerations. 24

We have argued here that by borrowing fromabroad through current account deficits, the U.S.may not ultimately suffer much increase in realinterest rates and may be able to maintain its pastinvestment levels. This does not mean that by bor­rowing from abroad our nation can entirely escapethe budget deficit burden, however. By borrowingfrom abroad the deficit's costs may be reduced(compared to the cost if we could not borrow), but asignificant burden is likely to remain.

As noted earlier, the deficits have temporarilyraised real interest rates. Even if this increase is notpermanent, housing and other interest-rate-sensitive

sectors of our economy certainly have suffered inthe interim. Furthermore, the high real value of thedollar brought about by increased real interest rateshas sharply reduced the demand for the output ofour traded-goods sectors.

Admittedly, by borrowing from abroad over thelast several years, the U.S. has probably been ableto maintain real interest rates at a lower level thanwould otherwise have been possible in the fa.ceofthe budget deficits. But this does not necessarilymean that the burden has been avoided-only that ithas been shifted from interest-sensitive to tradeablegoods industries. That is, in order for the U.S. toborrow from abroad (during the transition to thelong-term), our exports must shrink relative to ourimports, and this implies a reduction in the output ofour tradeable goods industries. In effect, budgetdeficits do "crowd-out" certain domestic industries,even in an open economy-and tradeable· goodsindustries may suffer as much or more than illterest­sensitive sectors.

Ultimately, the burden of a budget deficit is apt tobe manifest in lower (private) consumption for futuregenerations. In a closed economy, this burden comesabout as the lower investment resulting from thedeficit reduces the future capital stock, and hencefuture output available to meet the nation's needs.

An open economy like the U.S. may be able toavoid this reduction in its capital stock, and long­term output, by borrowing from abroad. However,the U.S. must still pay foreigners a portion of thatoutput to service its external debt Ineffed, theportion of our capital stock owned by U.S. citizenscan be expected to fall, even if the stock itself doesnot. Alternatively, the level of future U.S. outputmay not be reduced much, but the income from thatoutput earned by our citizens almost certainly willbe. In this sense, deficits do impose a long-termburden, one that is qualitatively the same as wouldoccur in a closed economy.

Despite these burdens, there can be little doubtthat our nation does benefit by its ability to draw onforeign funds to help finance our budget deficits. Asindicated earlier, our domestic real interest ratesalmost certainly will ultimately be lower as a result.Furthermore, to the extent that foreign borrowingallows the U.S. to maintain its investment and fu­ture output capacity, the productivity and wages ofour workers will be higher than they would be ifwecould not borrow from abroad. In this sense, thecurrent account deficits resulting from U.S. budgetdeficits are beneficial to our nation because. theyhelp to reduce, although not eliminate, the budget'sultimate burden.

IV. ConclusionOver the last several years, the U.S. has experi­

enced unprecedentedly high real interest rates, realdollar values, budget deficits and current accountdeficits. We have argued in this article that theseconditions are closely related and largely the resultof the increase in U.S. budget deficits that threatento remain at extraordinarily high levels for· manyyears.

Our budget deficits represent ademandforfundsthe government that must be met from an excess

of domestic saving over investment, or by borrowingfrom abroad, or both. In an open economy, anincreased budget deficit may be metpartiaIIy throughan increase in borrowing frorn abroad; its counter­part is an increase in the currenfaccountdeficit. Incontrast to the textbook closed economy case, thechannels transmitting the effects of budgetdeficitsto the open economy include exchange rates as well

20

as interest rates. This is particularly evident duringthe transition period before the currentaccount hasfully adjusted to a budget deficit. Initially, an in­creased budget deficit is likely to raise dornesticFe~1

interest rates which, in tum, raise the real e)(changerate. The higher real exchange rate then indyces .•••~current account deterioration that effects thetransf~r

of foreign saving to help finance the bUdgetdefi~.it.

A.fter several years, however, when the current af­count has fully adjusted to the bUdgetdeficiti~~\':'

the initial pressures on interesttatesa.re li~?l~ito

subside substantially, and real interestrates andthereal dollar should then fall backtowardlowerJevels.

In a growing world economy, ongoing US. cur­rent account deficits can in. principle b~ financedfrom foreign savings, and there is no/theoreticalreason and, in the absence of a shift in FederalReserve policy toward monetizing federal deficits,

/

few practical reasons why the United States couldnot borrow from foreigners for many years to come.The U. S. current account may therefore remain insubstantial defiCit as long as budget deficits of thepresent magnitlld", persist.

Our analysis has direct implications for policy­makers concefued about our growing trade defiCits.First, attelTIPts to eliminate our current account andtrade deficits by imposing trade barriers (for exam­ple, quotas, "voluntary" export agreements, tar­iffs, legislation of domestic content ratios for im­ports, and other measures), are likely to do moreharm than good to the economy. These measureswill raise costs to consumers and, by encouragingan inefficient and distorted alloction of our re­sources, may make U.S. industry less, not more,competitive in international markets. In addition, tothe extent that trade barriers are effective in redu­Cing our current account deficits and, hence, inreduCing foreign capital inflows, U.S. interest ratesare likely to be higher than would otherwise be thecase. This would both lower domestic private in­vestment and raise the overall cost of our budgetdefiCits.

Similarly, a more expansionary monetary policydesigned specifically to reduce real interest ratesand the value of the dollar in the foreign exchanges

21

also would most likely prove counter-productiveover the. longer term. In particular, a more. expan­sionaryU.S. monetary policy probably would. causethedoUar to depreCiate and eventually narrow theU. S. current and trade account. deficits. Not.onlywould this policy reversal undermine our hard~wongairis against inflation,· it could greatly undermineforeigners' willingness to lend to the U.S., aridhence reduce the extent to which we could financeour budget deficits by borrowing from them. Forthis reason, expansionary monetary policy couldultimately lead to higher real interest rates and lowerdomestic investment (greater crowding out) thanwewould otherwise suffer.

Thus, our analysis implies that if a reduction inour current account and trade defiCits is deemed animportant policy objective, the most effective andefficient measure for doing sO is through a majorreduction in the U.S. federal budget deficit. Only inthis way will our external deficits be reduced with­out creating either serious distortions in our liberaltrade environment oraresl.ltgenceofU.S. inflation.Conversely, in the absellceof a federal deficit re­duction, the benefits derived from continued for­eign savings inflows-the counterpart of our largecurrent account deficits-are likely to outweightheir costs.

[i - (pe-p)] + [se _(p*e_pe)J - [s- (p*-p)J

The left-hand expression and the first termon the right aresimply the foreign and domestic real interest rates, respec­tively; the second bracketed terms are the logarithms oftheexpected future and current real exchange rate. Theserelations apply, in principle, to all maturities. For furtherdiscussion, see Hutchison (1984).

7. This adjustment process is termed exchange rate "over­shooting". See Dornbusch (1976) for an original contribu­tion on overshooting in a simple monetaryrnodel of ex­change rate determination with sluggish price adjustment inthe goods market.

Note that the proportionality described in the text betweenthe exchange rate impact of an interest rate change and thematurity is strictly valid only for pure discount instruments.For coupon instruments, the impact is proportional to theduration rather than. the nominal maturity.

8. See Michael Keran, "Budget Deficits and Foreign Sav­ings," FRBSF Weekly Letter, July 6, 1984, for a discussionof the recent U.S. experience using an analytical frameworksimilar to that presented here.

9. The determinants of this lag can be fairly complex. Thelag could be very long if, for example, the policies under­lying the budget deficit were to raise substantially the returnto domestic investment. As explained later in the text, thiscould lead to an increase in the domestic share of the worldcapital stock to bring domestic and foreign returns to capitalback toward equality. Such a process is apt to take manyyears to be completed, however. The lag could also be veryshort, particularly if the policies generating the deficit di­rectly and immediately alter export and import demands.

10. The size of a nation's external debt in relation to itsGNP, and the share of foreign wealth that debt represents,can be related to the long-term current account/GNP ratio.To illustrate, suppose that the domestic and foreign econ­omies are growing at the same rate, "g" (allOWing theserates to be different does not significantly alter the conclu­sions). Then,

DIY = (CAIY)/g

where D/V is the long-term external debt (D) to GNP (V)ratio and CAIY is the long-term current account deficit(CA/GNP) ratio. This condition follows immediately fromthe observation that a constant D/V over time implies thatthe growth of external debt equals the growth rate of GNP.Similarly, it is easy to show that

C + I + G + (X-M) - R = C + S + T

i* = i + (se - s)

from which the relation (1) in the text follows immediately.

3. Note again that our analysis assumes a floating ex­change rate regime. The dynamic adjustment to a budgetdeficit under fixed exchange rates is verydifferent frorn thattraced in the text.

4. A large and growing literature exists on the relationsbetween budget deficits and real interestrates and output inclosed economies. Some have argued that deficits bear norelation to real interest rates in either a setting with less thanfull employment of resources or a full employment situation.This view is often termed the Ricardian equivalence prop­osition. Its central tenet is that the private sector is indif­ferent between tax- and deficit-finance of government ex­penditures, and that interest rates will not be affected by thedivision between the two forms of financing the government.(See J. Bisignano, 1984 for a complete discussion of thisissue). The discussion in the text assumes the receivedmacroeconomic theory holds, however, and that govern­ment budget deficits are likely to exert upward pressure onreal interest rates when the economy is at full employment.

5. A "risk premium" or equilibrium real interest differentialalso is included in this equation. We have subsumed thispremium within our "risk adjusted" real interest rate mea­sure for simplicity of exposition. See Hutchison (1984) for amore detailed discussion of risk premium determinants andreferences to the literature on the subject.

6. The text relation follows directly from the parity conditionfor nominal interest rates,

FOOTNOTES

1. A recent empirical study by Laney (1984) finds only in a i* _(p*e -p*)few cases a positive (statistically significant) link in thepostwar period between the external balance and the fiscalbalance for the major industrial economies. The empiricalinvestigation of the study showed a rnuchtighter linkage forthe smaller developing countries than for the industrialcountries, however, presumably. because of. the lack ofdomestic capital markets and inelastic private domesticsavings in developing nations.

2. In national income accounting terms, the value of na­tional output (V) equals the sum of private consumption (0)and investment (I), government expenditures (G), and ex­ports (X), less the amount spent on imports (M). Nationalincome-which equals the value of national output less nettransfers to abroad (R)-is divided into private consumptionand savings (S), and taxes (T). Hence,

This says that the foreign nominal interest rate (i* - ex­pressed as the percentage yield to maturity) must equal thedOrnestic nominal interest (1- for the sarne rnaturity andexpressed sirnilarly) plus the expeCted apprecia.tion (to ma­turity) of the nominal exchange r;:lte (s- expressed as thelogarithm of the foreign currency price of domestic cur­rency). Defining the logarithm of the domestic and foreignprice levels as p and p* respectively, while 'e' refers toexpected future values,

D/W* = (CAIY)(VIY*) -7- (W*IYX)g

where D/W* is the long-term external debt/foreign wealth(W*) ratio, VIY* is the ratio of home to foreign income (V*),and W*/v* is the foreign wealth/GNP ratio. Thus, for a givenconstant curient account/GNP ratio, there is a consta.ntdebt/GNP and debt/foreign wealth ratio in the long-run(admittedly, this conclusion is somewhat altered if domesticand foreign growth rates differ). Conversely, the external

22

debtlforeign wealth ratio can grow continually over timeonly if the CAIY is also growing, that is, does not level off.

11. Turnovsky (1976) and Sachs and Wyplosz (1984) ana­lyze budget policies for open economies in a static context,while. Hodrick (1980) considers their impact on growingeconomies. Comparison of their results demonstratesgraphically that the implication of a given bUdget policy candepend critically upon whether there is growth or not. Con­sider, for example, a policy that initially increases bothcurrent government expenditure and the deficit, but leavestaxes unchanged in both the short- and long-run. In a staticcontext, the government's budget (and the current account)must ultimately balance. This implies that in the long-run,the level of government expenditure mustactually fall fromits original level to allow the increased government interestpayments resulting from the initial deficit to be financed.When there is growth, expenditure need only fall back to itsoriginal level, s.ince an ongoing deficit to meet the increasedinterest payments is feasible. Thus, the nature of the gov­ernment's budget constraint is radically different in a growthcontext from that applying in a static context. For this reasonalone (and there are others), the implications of static mod­els canbe very misleading for actual experience.

12.••The risk that a government will prevent its citizens fromrepaying foreign debts, either by defaulting or denying themaccess to foreign currency, is also known as "political" risk.RiSks in lending to a country arising from unexpected ex­change rate changes, which we include among our macro­economic risks, are commonly known as "exchange risks".Note also that the risks associated with lending to a givencountry are not always the same as those incurred bylending in its currency. Sovereign risk generally does notdepend on the currency in which the loan is extended, whiletransfer (and exchange) risk does.

13. Applying relation (1) to "long-term" real interest rates(again expressed as percentage yield to maturity ratherthan on an annualized basis),

where qr. is the expected "long-run" real exchange rate.Suppose that there is no long-term secular trend in domes­tic relative to foreign prices and hence in the real exchangerate. Then, since (by definition) the real exchange rate mustultimately cOrne to equal its long-run value, the domesticlong-term real interest rate must settle to a level equal tothat of its foreign counterpart. (At this point,. short- andlong-term realinterest rates are equalized internationally).MOre generally, the domestic real interestrate must ulti­mately equal the foreign. real rate. plus the long-term, orsecular, rate of change of the real exchange rate. In eithercase,. it is evident that the equalization of domestic andforeign real·· interest rates-as conventionally defined interms of domestic and foreign price indices=is not depen­ctentuponttlespeedof arbitrage in financial markets,butonthe adjustmentof prices ingoods markets to their long-termvalues, a process which can take many years (see, forexample, the discussion in Niehans, 1984, Chapter 6).

14. The fact that the long-term real exchange rate impactof budget policies depends on how they affect (excess)demand for foreign vensus domestic gOQdS is discussed indetail in Sachs and Wyplosz (1984). See also Blanchard

23

and Dornbusch (1984). Hodrick's (1980) growth model ofbudget policies assumes foreign and domestic economiesproduce a single identical good and so does not considerreal exchange rate impacts.

15. For example, a recent study by Robert Feldman (1982)suggests that the nearly 30 percent real appreciation of thedollar over 1980-1983 reduced the U.S. trade balance byas much as $60 billion. See also Peter Hooper and RalphTryon (1984).

16. See, for example, a recent study by Peter Hooper(1983). See also Fielke (1984).

17. For a survey of this literature, see the recent U.S.Treasury study, The Effect of Deficits on the Prices ofFinancial Assets: Theory and EVidence (1984). This studyconcludes that there is no empirical support for a systematicreal interest rate-budget deficit link for the United States.Other researchers, however, do present evidence of asys­tematic linkage. (See Sinai and Rathjens, 1983 and Fried­man, 1982.) In addition, Hutchison and Pyle, in an accom­panying article in this ReView, find support for this linkageby looking at international evidence.

18. This view has been expressed in a recent article byArthur Laffer in the Los Angeles Times, (January24, 1984).

19. As this argument suggests, changes in fiscal policiescan have implications for interest rates, exchange rates,and the current account very similar to those discussed inthe text, even if they do not lead to a budget deficit. Theconverse is that our real interest rates and the dollar couldremain high even if our budget deficit were eliminated,provided that fiscal policy continued to encourage invest­ment in the U.S. (for example, if the deficit were reduced byraising taxes on consumption). Note also that to the extentthat investment incentives are responsible for recent de­velopments, the "adjustment" period required to bring ourinterest rates back to world levels may be very long, perhapsconsiderably longer than if budget deficits themselves werethe main cause. The reason is that changes in the after-taxyield to investment relative to abroad will ultimately have tobe offset by a "redistribution" of the world capital stocktoward the U.S. (to the point where the marginal revenue­product of capital falls enough to offset our more favorabletax treatment). This process could take many years.

20. Our current account's deficit's (CA) share of foreignsaving (S') can be written as:

CA/S' = (CAIY)(Y/Y') -i- (S'/Y')

where YIY' is the U.SJforeign GNP ratio, which is roughly55 percent, while S'/Y' is the foreign industrial country(net) saving rate, whose average is about 11 percent.These figures were derived from the OECD National Ac­counts: Main Aggregates, 1953-1982. They representaverages for 1979-1981.

In long-term "steady state", the current accountdeficitwill largely, if not entirely, consist of interest payments.toabroad while the budget deficit consists of interest paymentson the national debt. More precisely, if the long-term (U.S.)real interest and growth rates were the same, then thelong-term current account deficit would exactly equal thenet interest payments of the U.S. on its foreign debt,whilethe budget deficit equaled the interest payments On the

national debt. Thus, when the interest and growthrates arethe same, net exports of goods and services (that is, theportion of the current account excluding net interestpay­ments to abroad) must be in balance,. while. governmentexpenditures net of interest payments must equal govern­ment revenues. When the interest rate exceeds the growthrate, the current account and budgetdeficits are ultimatelyless than the respective interest payrnents. Governmentrevenues must then exceed non-interest expenditures andnet exports of goods and services must be in surplus in thesteady state. In general, then, a permanent deficit in thenon-interest portion of the budget is not possible. in thelong-run, at least not without some financing from moneycreation, unless the interest rate is below the growth rate.This has been pointed out by T Sargent and N. Wallace,"Some Unpleasant Monetarist Arithmetic," Federal Bank ofMinneapolis Quarterly Review. Fall, 1981, p. 1-17. Con­versely, when the interest rate is below the growth rate, thecurrent account deficit exceeds foreign interest payments,allowing the borrower to maintan along-term trade deficit.

21. In making this argument, we are relying as much,ormore, on theoretical plausibility as actual evidence. Thereremains considerable controversy about the degree of in­ternational capital mobility. While several studies suggestthere are country/currency risk premiums in short-terminterest rates for the major industrial nations (Meese andSingleton, 1980 and Hodrick and Hansen, 1980), there isvery little evidence to suggest they are very large, or sys­tematically related to a nation's external debt (Frankel,1982; Blanchard and Dornbush, 1984).

Several other studies have found that current account defi­cits have not historically contributed much to domestic in­vestment (Feldstein and Horoika, 1980 and Dooley and

Pennati, 1984). This finding is consistent with the hypo­thesis that international capital mobility is very low, but thispattern can also be explained in terms of other factors.Furthermore, there is some evidence that international cap­ital mobility has increased over the last decade; if so rela·tions among current account deficits and investment mayhave changed.

22. See McElhattan, 1984, as well as the references citedthere.

23. See David and Scadding, 1974.

24. In effect, fiscal policies represent choices in the waysociety's resources, present and future, will be allocated.Deficits can be viewed as one among many types of "taxpolicies" available to finance a given level of expenditure,present and future. They, in effect, represent taxes on future,rather than present, generations. Thus, deficits themselveshave no clear-cut welfare implications. Furthermore, the"costs" of deficits will generally be borne partly by thoseabroad (when the nation incurring the deficit can borrowfrom abroad). Again, as with any tax policy in an openeconomy, a portion of the burden may be "exported."

25. Recent work that suggests long-term interest ratesdiffer from the average of future shOrt-term rates (withoutarisk premium) has cast doubt on the simple rationalexpec­tations model of the term structure. Campbell and Shiller(1983) modify the "simple" expectations theory to includetime-varying risk premia, however, and appear better ableto reconcile actual movements in long-term and short·terminterest rates with the methodology. Nevertheless, underour assumed condition of perfect foresight (that is, no un­certainty), the simple expectations theory would hold as anarbitrage condition.

REFERENCES

Bisignano, Joseph, "CroWding Out and the Wealth Role ofGovernment Debt," unpublished mimeo, FederalReserve Bank of San Francisco, 1984.

Campbell, John and Robert Schiller, "A Simple Account ofthe Behavior of Long-Term Interest Rates,". NBERWorking Paper Series, No. 1203, September 1983..

David, Paul A. and John Scadding, "Private Savings, Ultra­rationality, Aggregation and Denison's Law," Journalof Political Economy, 82, Part 1, March/April 1974.

Dooley, Michael and Alessandro Penati, "Current AccountImbalances and Capital Forrnation in Industrial Coun­tries: 1949-1981," IMF Staff papers,July 1984.

Dornbusch, Rudiger, "Expectations. and Exchange RateDynamics," Journal of Political Economy, 1976, Vol.82, No. 61.

Feldman, Robert, "Dollar Appreciation, ForeignTrade, andthe U.S. Economy," Federal Reserve Bank of NewYork, Quarterly Review, Surnrner 1982.

Feldstein, Martin and Craig Horiaka, "Domestic Savingsand International Capital Flows," The EconomicJournal, June 1980.

Fielke, Norman, "Domestic and InternatiOnal Deficits,"New England Economic Review, May/June 1984.

24

Fleming, J. M., "Domestic Financial Policies Under Fixedand Under Floating Exchange Rates," IMF StaffPapers 9:369-79,1962.

Frankel, Jeffrey, ''A Test of Perfect Substitutability in theForeign Exchange Market," Southern EconomicJournal, 1982.

Friedman, Benjamin., "Interest Rate Implications for Fiscaland Monetary Policies: A Postscript on the GovernmentBudget Constraint," Journal of Money, Credit, andBanking, Vol. 14, No.3, (August 1982), pp, 407~412.

Hodrick, Robert, "Dynamic Effeetsof GdvernrnentPoliciesin a Open Economy," JME. April 1980,pp. 213--240.

and Lars Peter Hansen, "ForwardHatesasOp·timal Predictors of Future Exchange Rates," JPE, Oc­tober 1980.

Hooper, Peter, "Movements in the Dollar's Real ExchangeRate Over Ten Years of Floating: A Str(jcturaIAnaly~

sis," unpUblished working paper presented fortheThirdInternational Symposium on Forecasting, Philadelphia,Pennsylvania, June 7-8,1983.

and Ralph Tryon, "The Current Account oltheUnited States, Japan and Germany: A Cyclical Analy­sis," International Finance DiScuSsion Paper #236,

Bo~rdptGovernors of the Federal Reserve System,January 1984.

Hutchison, Michael, "lntervention,Deficit Financ~andRealExchange Rates: The Case of Japan," Federal Re­serve Bank of San Francisco, Economic Review,Winter 1984.

.._ ~~ and David Pyle, "The Real Interest Rate/BudgetDeficit Link: International Evidence, 1973-82," FederalReserve Bank of San Francisco,.EconomicRevieW,Fall 1984.

Keran, Michael, "Budget Deficits a.nd Foreign Savings,"Federal Reserve Bank of San Francisco, Weekly Let­ter, July 6, 1984.

Laffer, Arthur, "Beware of Curing the Imagined Ills," LosAngeles Times, Tuesday, January 24, 1984.

Laney, Leroy 0., "The Strong Dollar, the Current Account,and Federal Deficits: Cause and Effect," Federal Re­serve Bank of Dallas, Economic Review, January1984, pp. 1-14.

McElhattan, Rose. "Deficits vs. Investment," Weekly Let­ter, Federal Reserve Bank of San Francisco, June 29,1984.

25

Meese, Richa.rdand•Kenneth. Singleton, "Fiational E*pec­tation, Risk Premia and the Market for Spot and For­ward Exchal1ge," Discussion Paper #165 of the Inter­national Finance Division of the Federal ReserveBoard of Governors. July 1980.

Mundell. R.A., "TheA!:jpropriate Use of Monetary and Fis­cal Policy for Internal and External Stability," IMF StaffPapers 9:70-77, .• 1962.

Niehans, Jurg, International Monetary Economics (JohnHopkins University Press. Baltimore), 1984.

Sachs, Jeffrey and Charles Wyplosz, "Real Exchange RateEffects on Fiscal Policy," NBER Working Paper#1255, January 1984.

Sinai, Allen and· Peter Rathjens, "Deficits, Interest Ratesand the Economy," Data Resources Review, June1983, pp. 27-4.1.

Turnovsky, Stephen, "The Dynamics of Fiscal Policy in anOpen Economy," Journal of International Econom­ics, 1976, pp. 115~142.

U.S. Treasury Department. the Office of the Assistant Sec­retary for Economic Policy, "The Effect of Deficits onPrices of Financial Assets: Theory and Evidence," un­published monograph. January 1984.

Michael M. Hutchison* and DavidH.pyle**

There is a widespread beliefthat current and expectedfederal govern­ment credit demands are keeping U.S. real interest rates stubbornly highand may slow the speed and limit the duration ofthe economic recovery.To shed light on this debate, this study investigates the link betweenbudget deficits and real interest rates by "pooling" annual time seriesdata for the last decade across the seven major industrial countries. Theresults suggest that short-term real interest rates are systematically andpositively associated with central government budget deficits acrosscountries and across time.

There is a widespread belief that current andexpected federal government credit demands arekeeping U.S. interest rates stubbornly high and mayslow the speed and limit the duration of the eco­nomic recovery as it matures. This conventionalwisdom is generally supported by a body of macro­economic theory that posits a strong positive causallink between government budget deficits (or out­standing government debt) and real interest rates.

Nevertheless, there are theoretical challenges tothis proposition, and empirical support for it issketchy and largely based on indirect evidence de­rived from simulations of large scale econometricmodels. I Little empirical evidence of a direct linkrunning from budget deficits to interest rates hasbeen found. 2 In fact, the conclusion of a recentstudy by the U.S. Treasury (1984) was that " ... highdeficits have had virtually no relationship with highinterest rates ..." during the past two decades. 3

Other recent studies of the U.S. experience (for

* Economist. Federal Reserve Bank of San Fran­cisco.

** Booth Professor of Banking and Finance, Schoolof Business Administration, University of Cali­fornia, Berkeley, and Visiting Scholar, FederalReserve Bank of San Francisco

We would like to thank the Editorial Committee forhelpful comments and Mary Ellen Burton-Christieand Julia Lowell for excellent research assistance.

26

example, Evans, 1983, Motley, 1983, and Hoel­scher, 1983) also have failed to find a significantpositive link between U.S. budget deficits and in­terest rates.

Although considerable research has investigatedreal interest rate behavior and the relation betweenreal rates and budget deficits, very little of thisresearch has focused on countries outside the U. S.Extending the analysis to other countries could beuseful in several ways. For example, it could pro­vide information about the robustness of the resultfound for the U.S. Also, by extending the analysis,one can conduct joint tests for several countries atonce. This could result in more powerful statisticaltests of the deficit-interest rate link because moredata, that exhibit greater variation, can be exploited.

The latter consideration motivates the strategy ofthis paper. We pool annual time series data acrossthe seven major industrial countries (the U.S., theU.K., France, Japan, Italy, Canada and Germany)to investigate whether budget deficits are signifi­cantly positively associated with real interest rates..Pooling observations increases the variability of thedata, both over time and across countries, becauseof the diversity ofexperience with real interest ratesand budget deficits in the seven countries of thesample.

Using our pooled data sample, we regress short-­term real interest rates on budget deficits, holdingconstant money growth and a cyclical measure of

economic actlVlty. The results of our empiricalwork suggest that short-term real interest rates aresystematically and positively associated with cen­tral government budget deficits across countries andacross time. To testthe [obustnessofour surprising­ly strong. result, we test for the budgetdeficitltealinterest rate association in a variety of ways. Thebasic result does not appear particularly sensitive tothe choice of govemrnent deficit measure or moneygrowth proxy. However, the budget deficit/real in­terest rate linkage is weakened somewhat when thecyclical measure (a standardized unemploymentrate) is included as an explanatory variable.

On balance, this research provides empirical sup­port for the hypothesized positive linkage between

government budget deficits and real interest rates.This evidence is consistent with the view that bud­get deficits lead to high real rates of interest. To thebest of our knowledge, this is the first internationalevidence supporting the hypothesis.

The next section provides an outline of the meth~

odology we use to analyze the data. Section IIfollows with a description of the data and the empir­ical results from estimating the basic model. SectionIII extends the basic model and presents estimatesof various equations designed to test the robustnessof the deficit-interest rate relationship found in thepreceding section. The paper concludes with a briefsummary and some tentative policy implications.

I. MethodologyThe purpose of the research reported here is to

test the hypothesis that high short-term real interestrates (r) are positively associated with high buqgetdeficits (B), after controlling for other systematicinfluences on the real rate (Z):

The vector of other systematic variables, Z,should include all variables that are correlated withshort-term real interest rates. We limit Z, however,to a money growth variable (M), a cyclical variable-the standardized unemployment rate (U), andcountry-specific dummy variables (D).

In searching for the relevant systematic variablesto include in Z, we think of real interest rates asinfluenced by the demand and supply ofcredit in theeconomy. As money growth represents a net addi­tion to the supply of credit, it should be negativelyassociated with short-term real rates. The net cycli­cal private demand for credit, in contrast, will varywith the fluctuations in business activity-roughlyproxied by the unemployment rate variable. Thehigher the level of economic activity (the lower therate of umployment), the greater is the private de­mand for credit. However, higher business activityand income are also generally associated with great­er saving-increasing the net supply of credit. Thenet effect of the cyclical variable on the short-termreal interest rate depends on which factor dominates:the increase in private credit demand or privatecredit supply associated with expanding economicactivity.

r = feB, Z, JL) (I)

27

Thus, although we do not develop a completestructural model of real interest rate determination,our formulation of Equation (1) is consistent withboth the familiar IS-LM framework and a simpleloanable funds flow model of the bond and moneymarkets. 4 This body of macroeconomic theory pre­dicts a positive relation between budget deficits andshort-term real rates and a negative relation betweenmoney growth and short-term real rates. The ex­pected sign of unemployment is ambiguous, al­though most models generally predict a negativecorrelation with real interest rates.

The dummy variable for each country in the sam­ple is introduced to take into account some of theinstitutional and structural diversity, such as taxrates, non-homogeneous inflation measures, andpolitical instability among the industrial countries.These differences might explain persistent inter­national discrepancies in real interest rates (beyondthQse assQciated with money growth, governmentbudget deficits and cyclical variables). Taking noteof them is important because pooling data has themajor disadvantage of constraining the estimatedcoefficients in a model to be equal across countries.Introducing dummy variables is an attempt to cap­ture significant structural differences among coun­tries in the level of real interest rates. It does notcapture differences in their cyclical behavior. Nev­ertheless, it allows us to pool the sample and to use agreater degree of diversity in real interest rates andbudget deficits than would otherwise be the case.

The real interest equation that we have estimatedis:

r'i = bo + b l B'i + b2 M,;6

+ b1 U'i + 2: b1+; D,; + I-t'i (2), .. i=l"

where r'i = short-term real interest rate at time t incountry i

B'i government budget deficit (percent ofGNP) at time t in country i

M ti = money growth rate at time t in country i

U'i = unemployment rate at time t in country1

D ti = 1 for country i for all t,ootherwise

I-t,j = random error term

The estimation of Equation (2) using pooled dataimplies that the variation over time and acrosscountries in short-term real interest rates is not pure-

ly random but is due to structural (D), cyclical (U),and policy-determined (B and M) differences acrosscountri~s. In general, estimation on a country bycountry basis does not provide enough variation inth~budg~tdeficits to produce powerful tests of theireffec;t.oninterest rates. For example, there are. onlythree years between 1973-82 in which U.S. budgetdeficits were over 2.5 percent of GNP. Pooling dataover theseyencountries provides numerous obser­vations with deficits of this magnitude or greater.

To ()btainthe .added statistical power for discern­ing the effects of deficits, we constrain the policyand cyclical variables to have comparable effectsacross countries and ask, controlling for other vari­abl~s, does a higher deficit generally imply a higherreal interest rate for each country? Our interpreta­tion of OLS (Ordinary Least Squares) estimates ofthe coefficient b l in Equation (2) is that it reflectsthe correlation of short-term real rates with budgetdeficits, holding other relevant variables constant.

Notes: t-statistics in parenthesis; OLS regressions

Real Interest Rate Equations: PooledRegression for Seven Major Industrial

Countries, 1973-1982 Annual Observations

U.S.

Italy

(1) (2) (3)

.007 .031 .014(0.73) (2.83) (1.22)

.010 .008 .004(4.72) (4.09) (1.85)

-.003 -.003(-3.81) (4.11)

.007(3.15)

-.052 -.039 -.061(-3.94) (-3.16) (-4.50)

.016 -.005 -.022(-1.17) (-0.42) (- 1.74)

-.059 -.044 -.023(-4.06) (-3.22) (-1.56)

-.031 -.032 -.057(-2.31) (-2.59) (-4.08)

1.04 -.063 -.065(-5.69) (-3.19) (-3.55)

-0.03 -.034 -.057(-2.48) (-2.85) (-4.27)

.39 .51 .5870 70 70

.029 .027 .025

Table 1

France

Canada

Japan

Unemployment Rate

Dummy Variables:

U.K.

ObservationsStandard Error

Central GovernmentBudget Deficit

Nominal Money Growth

Constant

II. Empirical ResultsThe data used in our analysis are annual observa-

tions for the seven major industrial countries duringthe 1973-82 period. Only the recent decade waschosen because it represents a significant departurefrom the economic environment prevailing duringthe greater part of the 1950s and 1960s. Specifically,the breakdown of the Bretton Woods System offixed exchange rate parities, disruptive oil supplyshocks, rapid inflation and, in many countries, thelargest peacetime government deficits ever experi­enced, distinguish the last decade from the twopreceding. We believe therefore that the relationbetween budget deficits and real interest rates foundin the most recent data are likely to have the mostcurrent policy relevance.

Annual data are used because they are less likelyto be distorted by the transitory shocks that domi­nate short-term changes in some variables, particu­larly real interest rates. Annual data also appearpreferable because the timing between actual gov­ernment deficits and Treasury financing may notcorrespond very closely during shorter periods.Financial markets may adjust to new governmentdebt issues with some lag before reaching a newequilibrium situation. We hope to avoid problemsof this nature by using annual data.

The ex-post real short-term interest rate is thedependent variable in the regressions. Annual ex-

28

post real interest rateswere calculated from quarter­ly average rates (compounded annually) which, intum, were calculated by subtracting the actual CPIinflation rate over a quarter from the nominal inter­est rate for that quarter. Short-term interest rateswere employed in all countries to obtain a consistentandiriternatiollally comparable series of· market­determined interest rates. 5

Our empirical analysis focuses on the influenceofcentral government budget deficit on real interestrates. However, general government budget deficits(combining central and local governments) are em­ployed in several instances to test the robustness ofthe empirical results to the deficit measure choice.Deficits are measured as a percent of GNP to stan­dardize the figures for international comparison.(The central government budget deficit/gross sav-

ings ratio is also employed in several regressions.)Complete data definitions are given in the dataappendix.

The results from estimating Equation (2) are pre­sented in Table 1. This table also presents resultsfrom several formulations of the basic equation toprovide some insight into the stability of the esti­mated coefficients generally and, in particular, thestability ofthe budget deficit coefficient.

Coefficientestimates in the various formulationsof the model given in Columns (l)-(3) are, withoutexception, statistically significant with the theoreti­cally predicted signs. The nominal money growthcoefficient is not affected by the inclusion of thecyclical variable, unemployment. The budget defi­cit coefficient is larger, however, when unemploy­ment is excluded from the model. The unemploy-

Table 2

Real Interest Rate Equations:Real Money Growth and Cyclical Money Growth;Pooled Sample: 1973-1982 Annual Observations

Real MoneyGrowth Included

Cyclical MoneyGrowth Included

Constant

Central Government Deficit

Real Money Growth

Cyclical Money Growth

Unemployment

Dummy Variables:

U.K.

France

Japan

Canada

Italy

U.S.

R2ObservationsStandard Error

.077(0.69)

.010(4.62)

.0002(0.30)

-.051(-3.72)

-.016(-1.14)

-.059(-4.00)

-.029(-2.11)

-.103(-5.57)

-.032(-2.32)

.3970.029

-.009(-0.85)

.006(2.43)

.0001(0.01)

.007(2.72)

-.073(-4.75)

-.032(-2.26)

.038(-2.37)

.055(-3.38)

-.106(-6.01)

-.055(-3.52)

.4570

.028

.005(0.45)

.011(4.89)

-.001(-1.73)

.052(-3.94)

.013(-0.95)

-.062(-4.27)

-.031(-2.32)

.105(-5.83)

.031(-2.36)

.4270.029

.013(-1.22)

.006(2.54)

-.001(-2.06)

.007(2.99)

-.074(-5.13)

-.030(-2.17)

.039(-2.56)

.057(-3.74)

-.107(-6.37)

-.054(-3.73)

.4970.027

Notes: t-statistics in parenthesis; OLS regressions

29

ment coefficient is statistically significant and posi­tive. This suggests that a cyclical downturn (proxiedby an increase in the unemployment rate and givenunchanged budget deficits and money growth rates)is associated, on average, with. a risein real interestrates. 6

These results lend support to the commonly heldview that large budget deficits( as a percent ofGNP)are one factor causing high real. short..term interestrates. In particular, these estimates suggest that aone-percentage point increase inthe budget deficit!GNP ratio over a one-year period is associated withan average real interest rate increase of between

40-100 basis points. Rl1pid nominal money growth,on the other hand, is associated with lower realinterest rates. Aone-percentl1ge point increase inthe annual rate of narrow money growth is associ­ated with an average real interest rate decline ofapproximately 30 basispoint~.This result parallelsthose found earlier by Mishkin (1984) ina study ofreal interest rates in the Euro-deposit market usingquarterly data over the· 1967-Hto 1979-Ilperiod.

When combinedwith the trends in budget deficitsand. in nominal money·growth, these empirical re­sults imply that the sharp increase in short-term realinterest rates in the U.S. between 1980-1982 is

Table 3

Real Interest Rate Equationswith General Government

Deficits and the Central Government Deficit­Gross Savings Ratio as Independent Variables

General GovernmentBudgetDeficits

Central GovernmentBudget Deficitsl

Gross Savings Ratio

Constant .032 .012 -.021 .034 .016 .013(2.37) (0.95) (-1.79) (3.09) (1.33) (-1.10)

General Government Budget Deficit .005 .001 .004(2.31) (0.64) (1.75)

Central Government Deficit/Savings Ratio .002 .0007 .0009(3.84) (1.37) (1.89)

Nominal Money Growth -.003 .003 .003 -.003(-3.74) (-4.17) (-3.95) (-4.23)

Cyclical Money Growth -.002 -.001(-2.09) (-1.95)

Unemployment .009 .009 .008 .008(4.23) (4.07) (3.18) (3.05)

Dummy Variables:

U.K. -.039 -.066 -.081 -.042 -.063 .078(-2.89) (-4.82) (-5.60) (-3.29) (-4.62) (-5.32)

France -.003 -.027 .032 -.006 -.024 -.034(-0.23) (-1.99) (-2.22) (-0.50) (-1.85) (-2.37)

Japan -.023 -.009 -.021 .031 -.014 -.027(-1.69) (-0.69) (-1.60) (-2.38) (-1.09) (-1.95)

Canada -.018 -.059 -.055 -.032 -.058 -.059(-1.37) (-3.89) (-3.31) (-2.56) (-4.08) (-3.73)

Italy .049 -.056 107 -.057 -.059 -.101(-2.05) (-2.65) (-5.54) (-2.93) (-3.28) (-5.97)

U.S. .022 -.059 -.052 -.037 -.059 -.058(-1.59) (-3.98) (-3.17) (-3.03) (-4.43) (-3.91)

R2 .42 .56 .47 .49 .57 .47Observations 70 70 70 70 70 70Standard Error .029 .026 .028 .027 .025 .028

Notes: t-statistics in parenthesis; OLS regressions

30

partly attributable to slower money growth (120basis points), but that the rise in federal budgetdeficits over tllep¢riod probably played a larger role(100-250 basis points).

Experiments using alternative explanatory vari­able definitiorisare reported in Tables 2 and 3. Thefocus of these experiments is to determine the sensi­tivity of the budget coefficient to alternative specifi­cations. For the regressions reported in Table 2,alternative money supply definitions, real moneygrowth (nominal money growth less actual infla­tion) and a measure of cyclical money growth (cur­rent money growth less the weighted average ofmoney growth during the preceding three years)were used. 7 In the macroeconomics literature, thesemoney growth variables are often offered as alterna­tives in real interest rate equations to the nominalmoney growth variable. The use of either of thesealternative money supply variables increases themagnitude and significance of the central govern­ment budget deficit coefficient as compared to thosereported in Table I for nominal money growth. Thereal money growth coefficients are statistically in­significant and of the wrong sign, however.

Other experiments were conducted to addressquestions about the appropriate budget deficit mea­sure, for example, whether it should be limited tothe central government or should also include defi­cits due to local governments and social securityfunds. The first three columns of Table 3 give theresults from substituting general government defi­cits (central government plus local governmentsand national social security funds) scaled by GNPfor the central government deficit variable used in

the previous regressions. Somewhat surprisingly,the results for general government deficits in Table3 are. similar to the results using the central govern­ment budget deficit/GNP ratio presented in Table Iand Table 2. The coefficient fOf}hegeneralgov­ernment deficit/GNP ratio is significant in two ofthe three cases , and the effect of the deficit on realinterest rates is similar to that when only centralgovernment budget deficits are used. However,when both nominal money growth and unemploy­ment are included in the estimation, the coefficientfortlle general government budget deficit/GNP ratiofalls off in magnitude and significance.

The scaling of budget deficits by GNP rather thansome other aggregate may also be questioned. Col­umns (4)-(6) of Table 3 report on the use of centralgovernment budget deficits scaled by gross nationalsavings instead of GNP. This alternative deficitvariable measures the degree to which savings in aneconomy are absorbed by central government defi­cits and might provide a better indicator of thepressure exerted on real interest rates. 8 These resultsalso generally support the earlier findings. The cen­tral government deficit/gross savings coefficient hasthe anticipated sign in each equation, but, again, itis not significant in the equation that includesunemployment.

In all of the experiments reported in Table 3, themagnitudes and significance levels of the moneysupply and unemployment variables are similar tothose reported in Tables I and 2. On balance, theresults in Table 3 suggest that there is significantpositive correlation between deficits and short-termreal interest rates regardless of how deficits aremeasured.

III. ExtensionsApart from measurement issues, the simple em­

pirical tests we have reported may be criticized onseveral levels. One criticism concerns the appropri­ateness of pooling the data, and thereby constrain­ing the slope coefficients of the model to be equalacross the seven countries. A second criticism couldquestion the homogeneity of the data across theperiod ofestimation. Some economists have arguedthat major structural changes in the world economyhave occurred since 1979, potentially changing thelinkages between budget deficits and real interestrates. These potential objections to the methodology

31

behind the estimates reported in Table I are dis­cussed in tum below. 9

Pooling the DataPooling cross-section and time-series data impli­

citly assumes that all observations come from thesame. population regardless of country. As a test ofthe appropriateness of this assumption, we allowedthe coefficients of the fully constrained model, thatis, the model with a single intercept term (no coun­try dummy variables), money growth variable andcentral government deficit variable, to vary across

Notes: I-statistics in parenthesis; OLS regressions

D is a dummy variable and equals zero for the 1973-1979period and unity for the 1980-1982 period.

Table 4

Real.lnterest R.ate Equations:Tests of Structural Change

Between 1973·1979 and 1980-1982

confidence of better than 90 percent (the F statisticequals 2.05), that a structural shift did occur be­tween the two sets of parameter coefficients esti­mated<over the 1973-1979 (Column 1) and 1980­1982 (Column 2) periods. Somewhat surprisingly,the correlation between money growth and real in­terest rates becomes less significant in the latterperiod. In contrast, the positive correlation betweenreal interest rates and budget deficits increases sub­stantially; In addition, the structural change test forthe budget deficit coefficient alone during the latterperiod (Column 3 in Table 4) suggests that it hasincreased significantly, from .006 in 1973-1979 to.01(.006 +.004) in 1980-1982. Hence, while theevidence suggests that significant structural changes

Constant .026 .016 .026( 1.89) (0.85) (2.36)

Central Government .006 .013 .006Bu(iget Deficit (2.77) (2.42) (2.94)

Money Growth -.002 .0006 -.002(-2.15) (0.25) (-2.42)

Dummy Variables:

U.K. -.015 -.031 -.043( -3.81) (-1.36) (-3.49)

France .054 .021 .009( -1.08) (-0.73) (-.80)

Japan -.047 -.055 -.046(-3.07) (-2.03) (-3.43)

Canada -.032 -.036 -.031( -2.28) (-1.69) (-2.64)

Italy -.059 .137 .034(-2.89) (-2.52) (-2.87)

U.S. -.041 -.029 -.034(-2.94) (-1.31) (-2.87)

D* Central Government .004Budget Deficit (2.24)

R2 .47 .49 .54Observations 49 21 70Standard Error .026 .023 .026

1973­1982

1980­1982

1973­1979

countries .. Denoting the model with no restrictionsacross countries (intercept terms, money growthslope coefficients and central government deficitcoefficients are all allowed to vary across countries)asthe "expanded model," the relevant F-statisticmeasuring.thesignificance of the reduction insquared errors between the expanded model and thefully constrained model equals 1.43. The criticallev~loftheF-statistic at the 5-percent level ofconfidencefor the test is k92. This result suggeststhatpooling may be appropriate, since the F-valueis substantially below the value needed to reject (atth~5-percent level) the null hypothesis that all ob­servations come from the same population. Becausethistest is also unnecessarily strong, since our modelformulations in Table 1 also include unrestrictedcountrY-specific intercept terms, one can be evenmore confident of the result.

Tests for Structural ChangeThe international evidence reported here contrasts

markedly with numerous domestic studies that sug­gest very little association between budget deficitsand real interest rates. In fact, the lack of strongempirical evidence supporting almost any reason­able hypothesis attempting to explain the high levelsof real interest rates since 1979 have led some eco­nomists to suggest that a major structural change(presumably unquantifiable) has occurred in theprocess generating real interest rates.

To shed some light on the issue, we split oursample into two periods, 1973-1979 and 1980-1982,and estimated the real interest rate equation for bothsubperiods. The year 1980 is chosen as the breakingpoint of the sample because of the rapid run-up inreal interest rates that began in that year, and be­cause it is the first year following the October 1979policy shift by the Federal Reserve toward monetaryaggregate targeting. In another structural changetest, the budget deficit coefficient alone is allowedto vary between the two sub-periods, while themoney growth and country intercept variables areconstrained to be equal over the full 1973-82 sample.This allowed us to compare the significance of thebudget deficit-realinterest rate link in the two peri­ods and, in particular, to test whether the significantpositive •association noted above has becomestronger since 1980.

The estimated equations are reported in Table 4.The results ofa Chow test suggests, with a degree of

32

have occurred in the world economy in recent years,it appears that the significant positive correlation

between budget deficits and real interest rates re­mained, and may have grown stronger. 10

IV. ConclusionThispaperpresents somesimple tests ofthehy­

pothesis thathigh real interest rates in recent yearsare significantlycolTelated with large centralgov­ernTent budgetdeficits.We look at internationaldatato explore this linkage. Specifically,we poolannual observations over theiastdecadefor theseven major industriaL countries •and regress realillterest rateS9IlceIlt:r;l1 governrnent deficits,holdingconstant money growth, standardized unemploy­ment rates and country-specific dummy variables.

The results generally support the hypothesis. Wefind that a statistically significant positive relationgenerally holds between real interest rates and def­icits, irrespective of included variables, moneygrowth measures and deficit measures. Althoughthere are SOme. ~xceptions to this conclusion, thepositive deficit-real interest rate correlation is suffi­ciently stable in a variety of model formulations togive us confidence in the robustness of this empiri­cal result.

The policy implications from these results shouldbe drawn cautiously. Strictly speaking, the single

DATA APPENDIX

Data Sources: International Monetary Fund, InternationalFinancial Statistics (IFS); Organization for Economic Co­operation and Development (OECD-1), Occasional Stud·ies, June, 1983; Organization for Economic Cooperationand Development (OEC;D-2); economic Outloqk, Decem­ber1983.

1. Central anQ QeneralQovernment deficits as a percent ofGOP are from OECD-1.

2. Nominal mcmeygrowth rates are from IFS, line 34x.3. CPiinflation rates are from IFS, line64x.4. Domestic short term interest rates data are from:

France, call money rate, IFS line 60bU.S., Treasury bill rate,lFS line 60cGermany,p<:illmoneyrate,IFS line 60bU.K.,Treasury bill. rate,lFs line60cJ<:ip<:in, callrnoneyrate,.IFS line 60bItaly, call moneYrate, IFSline60cC<:inada,Tre<:islJry bill rate,lFSline 60c

5. Unemployment rates are fromOECf).2.6. Countryspecific<dummy variables equal 1 for named

country,06therwise.

33

equatioh tnethodology employed here qot}s notal­low us to determine the causal linkage betweenbudget deficits and real interest rates. Nevertheless,our evidence is clearlyco~sistentWiththe hypothe­sis that sizellble central~ovemment budget (ieficitsin the world's major economies maybe. animpQrtantfactor holding up teal interest rates in recentyearS.

The results of our. empirical analysis are •alsoconsistent with the hypothesis thattherecentslow­down in monetarY growth may also bear part oftheresponsibility for current high real interest nlt~s. Inparticular, the empirical results suggest that thesharp increase in U. S. real interest rates between1980-198:2 is partly attributable to slower moneygrowth (120 basis points), but that the rise infederalbudget deficits over the same period likely played alarger role 000-250 basis points). Moreover, inlight of the evidence gleaned from splitting thesample into the pre- and post-1980 period, itap­pears likely that large central government budgetdeficits are playing an increasingly important role inmaintaining real interest rates at their present le"els.

FOOTNOTES

1. Simulations of the familiar MIT-Penn-SSRC economet­ric model, for example, suggest a substantial increase inshort-term interest rates given a fiscal stimulus (see Fried­man,. 1982). The Federal Reserve Board's mUlti-country(MQfv1) modeLalso predicts a rise in interest rates in re­SPOnse to a fiscal stimulus (see Haas and Symansky 1983).Both of these models are large-scale structural models ofthe economy.

2. Reduced form estimates of a direct link between budget<ieficitsand interest rates have generally been unsuccess­ful.Recent work has investigated various aspects of apotential reduced form linkage. Motley (1983) considers theimpact otseveral measures of budget deficits on real inter­estr~tEll) and finds little correlation. Hoelscher (1983), onthe other hand, investigates the impact of bUdget deficitsduring the post-war period on nominal interest rates andfinds nO statistically significant relationship. Evans (1983)takes a different perspective and investigates war-timeexperiences, again failing to find the hypothesized deficit­interest rate relation. One recent empirical study (Sinai andRathjens, 1983) purports to have found a significant link,however, on interest rates arising from real governmentdeficits per capita in the U.S. Feldstein and Eckstein (1970),in an earlier study, also estimated a significant effect for realper capita federal government debt on interest rates. Nev­ertheless, a significant body of literature has failed to findany systematic support for a strong deficit-interest rate link.

3. U.S. Department of Treasury (1984).

4. Hoelscher (1983) derives this model and discusses itsrelationship to the standard IS-LM model. This reducedform equation is consistent with various theoretical frame­works. The important point to note, however, is that thismodel is a flow model rather than a stock model of interestrate determination.

5. Mishkin (1984) discusses the assumptions implicit inusing the ex-post real interest rate as a proxy for the expect­ed· (ex ante) real interest rate. Basically, this approachassumes that markets are efficient, that is, all transactorsutilize all available information in forming their expectationsabout future inflation. The assumption that ex-post realinterest rates are good proxies for ex-ante expected realinterest rates is quite strong, perhaps unnecessarily so. Forexample, however one proxies inflationary expectations,the assumption that the inflationary expectations are ab­sorbed in nominal interest rates on a one-for-one basisplaces an additional constraint on the estimation. This con­straint can be relaxed by re-arranging equation (2) to makerealized inflation the dependent variable and the nominalinterest rate an additional explanatory variable. Thischange in the estimated equation resulted in modest in­creases in the magnitude and the significance of the budgetdeficit coefficient. Specifically, the regressions reported incolumns 2 and 3 of Table 1 were re-run with realized infla­tion as the dependent variable and nominal interest rates asan additional explanatory variable. Except for the anticipat­ed change in sign, the coefficients for money growth and forunemployment are virtually unchanged in either magnitudeor statistical significance. Regressing realized inflation onthe nominal interest rate, the central government deficit,nominal money growth, and the country dummies results ina coefficient (t-statistic) for the central government budget

34

deficit of ~0.010 (-4.28) as compared with 0.008 (4.09) inthe original formulation. With unemployment included as anl:lxp1anator)' variable, the comparison is -0.006 (- 2.10) VS.

0.004(1.85).

6.Ul1Elrnploymentwas included in the estimation equationtoattempttocQntrolJactors associated with businesscyclefluctuations. Unemployment, however, has a significantrising trend component. To address this.problem, we con­structed a "CYClical" unemployment variable as the devia­tiOn pf current unemployment from its past weighted average(three previous years; weights equal to .5 for the first yearand.3and.2, respectively, for the second and third years)and used this constructed variable in the equation in column(3) ofTabll;ll .• The results of the regression provided strongsupport for the hypothetical budget deficit/real interest ratelink; estimated coefficient values (t-statistics) equal .007(3.12) for the central govemment budget deficit, - .003(-3.89) for nominal rnoney growth and .004 (1.12) for theconstructed cyclical unemployment variable.

7. The weights on lagged money growth are .5 for the firstyear, .3 for the second year and .2 for the third year.

8. Capio, et al (1983) discuss the problems involved withusing deficit-savings ratios as indicators of interest pres­sure in both the closed and open economy context. A majorcriticism of deficit-savings ratios posed by the paper is thatdeficit-savings ratios are not exogenous indicators and areaffected differently across countries given similar exoge­nous Shocks to. government deficits.

9. A third potential criticism is the question of simultaneityin the estimation of the real interest rate equations. Althoughwe do not suggest that the observed correlation betweenbudget deficits and real interest rates necessarily impliescausation, the empirical results may be interpreted this wayby some. Critics of this latter view may therefore raise thesimultaneity issue. Specifically, it may be argued that anincrease in real interest rates depresses economic activitywhich, in turn, causes government budget deficits to rise.Thus, the budget deficit/real interest rate correlation maybepicking up a reverse causation running from real rates tobudget deficits.

To .address this issue, we estimated equation (2) in the textuS.ir)gsl;lyer~lins~(umental variables procedures. The in­strumental variable estimates also suggest a positive bud­get deficit/real interest rate association, but do so lessst(Qllgly thanttle >OLS results. In. Particular, the budgetdeficit coefficient is statisticall¥ significant at the 10% levelwhen the unemployment variable is excluded, but drops offin significance when it is included in the instrumental vari­able estimates; We are not satisfied with our instrumentalvariable results and do notreport them in the text becausewe could not find an appropriate exogenous instrument forthe budget deficitvariable for each country.

In addition, itShould be p()inted out that a similar simultan­eity argument is most commonly made to explain why bud­get deficits may be negatively associated with real interestrates. Namely, an exogenous fall in real income may simul­taneously cause both a fall in real interest rates and anincrease in government budgetdeficits. Proponents of thisview therefore .a(gue that "structural" budget deficits (ahypothetical estimate of the budget deficit based on a full-

employmentoutputlevel in theeconomy)are positivelyassociated with reCiI interest rates,even though observeddeficits may bEl.l1egativ~ly associated with re(jJinterestrates. This commonly held view suggests that our esti­mates of a positive correlatiol1 between unadjusted budgetdeficits and real interest rates have a negative bias (forexample, should have a stronger positive correlation than isreported),. rather than the positive bias discussed in thistext.

10.. This result is weakened, however, when.the nominalmoney growth coefficient is allowed to.vary together withthe budget deficit coefficient during the latter period whilethe other coefficient values are. constrained to be equal.Nevertheless, in this .instance the evidence also suggeststhat the influence of budget deficits on real rates of interesthas increased in recent years.

35

REFERENCES1. Gerald Capio, etal, "Deficit-Saving Rlitios As Indical:ors

of Interest .Rate Pressure: A 9ollE!(:tion of Notes,"Board OfGovernors of the Federal Reserve System,lntern;itiollal F'inan~e.Dis~ussionPapers, No. 234,December 1983.

2. George Demopoulos, et ai, "Central Bank Policy andth~F'inancin9ofGovernment Budget Deficits: ACross'-Country Comparison," Economic Papers ofthe Commission of the European Communities No. 19,September 1983.

3. Paul Evans, "Do Large. Deficits Produce High InterestRates?"· Unpublished Working Paper, Stanford Uni­versity,October1983.

4. Martin S. Feldstein and Otto Eckstein, "The Fundamen­tal Determinants of the .ll1teresl Rate," Review ofEconomics and Statistics, November 1970, pp.363-375.

5. Benjamin M. Friedman, "Interest Rate Implications forFiscal and MonetaryPolicies: A Postscript on the Gov­ernment Budget Deficit," Journal of Money, Creditand Banking, Vol. XIV, No.3., August 1982.

6. Richard A Haas and Steven A. Symansky, "AssessingDynamic Properties ofthe MCM: A Simulation Ap­proach," Board of Governors of the Federal ReserveSystem, International Finance Discussion PapersNo. 214, January 1983.

7. Gregory P. Hoelscher, "Federal Borrowing and Short­Term Interest Rates," Southern Journal of Econom­i~s, Vol. 50. No.2, October 1983, pp. 319-333..

8. Frederic S. Mishkin, ''The Real Interest Rate: A Multi­Country Empirical Study," Forthcoming. The CanadianJournal of Economics (1984).

9. Brian Motley, "Real Interest Rates, Money and Govern­ment Deficits," Federal Reserve Bank of San FranciscoE~()nomic Re"iew, No.3, Summer 1983, pp. 3H5.

10. Robert w.R. Price and Jei:J,n-Claude Chouraqui, "PublicSector Deficits: Problems and Policy Implications,"OECD Economic· Outlook: Occasional Studies,June 1983, pp. 13-44.

11. Allen Sinai and Peter Rathjens, "Deficits, InterestRates and the Economy," Data Resources U.S. Re­view, June 1983, pp 1.27-1.41

12. U.S. Treasury Department, the Office of the AssistantSecretary for Economic Policy, "The Effects of Deficitson Prices of Financial Assets: Theory and Evidence."Monograph, January 1984.

nnouneemen

John P. Jl.ldd*

The positive association between long-term interest rates and M1"surprises" typically is ascribed to changes in inflation expectationsinduced by the unanticipated money supply changes. The empirical testsin this paper use a. term structure framework for explaining long-terminterest rates to isolate more carefully this effect. One of the surprisingresults is that the expectations effect does not show up in the pre-October1979 data, the period in which it is generally assumed that the Fed'sprocedures for controlling money were less well designed to fight infla­tion. An hypothesis that the presence oflarge government budget deficitsin the post-1979 period may account for the presence of a significantexpectations effect then does not appear to be borne out by the data.

In recent years, many financial and monetaryeconomists have offered explanations for the strongresponses of interest rates and exchange rates to theFederal Reserve's weekly Ml-announcements; Forexample, when an announced increase in MI islarger than expected by the market (or a decrease issmaller than expected), short- and long-term interestrates generally increase on the following day, andthe dollar appreciates in the foreign exchange mar­ket. The opposite movements generally are observedwhen MI comes in below market expectations.

This phenomenon probably arises because themarket perceives that the Fed attempts to exercisecontrol over MI, but that it does so somewhat cau­tiously. That is, in the short-run, the Fed is per­ceived as attempting to offset some, but not all, ofthe deviations of Ml from target. The responses ofshort-term interest rates and exchange rates there­fore probably reflect a policy anticipations effect:when Ml increases more than anticipated, the mar­ket expects the Fed to tighten monetary policy tem­porarily, which will raise real short-term interestrates and exchange rates. Responses of long-termrates have been interpreted primarily as reflectingchanges in inflation expectations: the market ex-

* Research Officer. David Murray and David Taylorprovided research assistance.

36

pects the Fed to react against a money surprise, butnot to react strongly enough to prevent some in­crease in inflation in the future. This phenomenon iscalled the expected inflation effect.

Given these effects, it is nevertheless puzzlingthat long-term bond rates have responded as sharplyas they have in the past five years to money supplyannouncements, and that they have responded moresharply since October 1979 than before. Since theFed changed its operating procedures in October1979 to enhance control over M1to attain better theobjective of gradually bringing inflation down, itwould make more sense if bond rates respondedmore sharply to Ml-surprises in the earlier periodthan in the later one. In this paper, we attempt tosolve this puzzle in two separate ways. First, insteadof analyzing the behavior of long-term bond rates,as most previous studies have done, we examineshort-term interest rates expected to prevail in thedistant future. For the reasons discussed below,changes in long-term rates can be difficult to inter­pret because they reflect changes in short-term realrates as well as expected inflation. However, theexpected future short-term rates examined in thisstudy should provide "cleaner" estimates of theresponses to expected inflation, and therefore mayhelp solve the puzzle described above.

Second, we examine the possibility that the exis-

tence of "large" expected future structural budgetdeficits in recent years might have affected the sizeof responses to money supply announcements. Theargument is that when current and future federaldeficits are high, positive Ml-surprises may tend tointensify fears that the Fed might monetize part ofthe government debt associated with the deficits.Conversely, negative Ml-surprises could reduceconcerns about monetization. Thus, the change infiscal policy regimes that occurred in 1981, whenexpected future deficits apparently became large,could have affected the responses of interest rates toMI-surprises.

The tests in this paper support the conclusion thatthere is both a policy anticipation effect and aninflation expectation effect operating simultane­ously. Their existence confirms the idea that themarket believes the Fed pursues its monetary con­trol objectives somewhat cautiously. Moreover, thetests provide some evidence that a stronger policyanticipations effect is associated with a weaker in­flation expectation effect. Finally, the results sug­gest that expected inflation effects are significantfor only about seven years into the future, and thatthese effects, together with movements in current

and near-term real interest· rates, account for theobserved large responses of 30-year bond rates.

The major remaining puzzle is why there was nota significant expected inflation effect prior to theFed's anti-inflation policy that began in October1979. As noted above, this paper examines thehypothesis that this apparent inconsistency might berelated to the change in fiscal policy regimes. in1981. It is possible that greater fears ofmonetizationin the latter period may have caused the inflationexpectations effect to be larger. However, tests ofthis hypothesis met with only mixed success, andthis puzzle will have to be solved by future research.

Section I reviews the literature on money supplyannouncements, and points out the apparently puz­zling behavior oflong-term interest rates. Section IIpresents empirical estimates of the responses ofshort-term spot rates and several forward interestrates to Ml-surprises in three monetary control re­gimes: September 1977-0ctober 1979, October1979-0ctober 1982, and October 1982-February1984. Section III tests for effects of the change infiscal policy regimes in mid-198I. Conclusions arepresented in Section IV.

I. The Money Supply Announcement PuzzleIn recent years, economists have produced a

plethora of journal articles on the responses of vari­ous financial asset prices to the Federal Reserve'sweekly announcements of M 1. 1 Interest in this sub­ject became intense following the Federal Reserve'schange in operating procedures in October 1979.Prior to that date, the Federal Reserve had attemptedto control money over periods of several quartersthrough very gradual changes in the Federal fundsrate. 2 Under the new procedures, which used non­borrowed reserves as the instrument of monetarycontrol, the Fed permitted short-term interest ratesto vary in the short-run more than they had previ­ously.3 Greater interest rate volatility was consid­ered necessary to achieve greater control over themonetary aggregates in the short-run. Moreover,for most of the period up to the fall of 1982, MI wasgiven the most weight in monetary policy decisions.

Coincident with the change in Federal Reserveoperating procedures, interest rates and foreign ex­change rates began to respond strongly to the Fed's

37

weekly announcement of the most recent weeklyM 1 figure. Actually, the reactions were systematiconly when changes in M1differed from the changeexpected by the market. Thus, changes in M1anti­cipated by the market seemed to induce no systema­tic response in asset yields, presumably because theresponses previously had been incorporated intoyields.

A large number of studies have estimated theresponses of yields to unanticipated changes in M1,both before and after the October 1979 change in theoperating procedures of the Fed. These studies usedsimilar econometric techniques in their tests. Theexplanatory variable in the regressions was the"surprise" in the change in weekly Ml-that is,the change in the actual M1 announced by the Fedminus the change in expected M I. The latter vari­able is measured as the median value of a set offorecasts by money market economists surveyedand recorded by Money Market Services, Inc. Thissurvey has been conducted from September 1977 to

the present. The dependent variables in previousstudies include changes in a variety of long- andshort-term interest rates and exchange rates, wherethe change is measured from the end of the day oftheM I announcement to the following day.

Although different studies have used somewhatdifferent sample periods, they have obtained verysimilar results. Prior to October 1979, the responsesof short-term interest rates, long-term interest ratesand dollar exchange rates were very small, andgenerally (though not always) statistically insignifi­cant. For example, using data from January 1978 toOctober 1979, Cornell (1983) found that, in re­sponse to a I-percent positive M I-surprise, thethree-month Treasury bill rate rose by (statisticallyinsignificant) 2 basis points, and the 30-year Trea­sury bond rate fell by (statistically insignificant) 0.4basis point. A very small response also was foundfor the dollar price of the German mark. FromOctober 1979-December 1981, the response be­came highly significant. A I-percent positive M 1­surprise was associated with a (statistically signifi­cant) 30-basis point increase in the three-monthTreasury bill rate, a 15-basis-point increase in the30-year Treasury bond rate, and an appreciation ofthe exchange rate.

Alternative Theoretical ExplanationsA number of theories would predict responses of

asset yields to MI-surprises. To understand thesechannels of influence, it is useful to consider theFisher equation (I). This equation states that thenominal interest rate (i) of a particular holding peri­od can be decomposed into the real interest rate (r)of the same holding period and the expected rate ofint1ation over that period (pC). Both the real andint1ation components of the nominal rate depend onexpectations about the future. M I announcementscan cause changes in interest rates by altering thoseexpectations.

r + pC (I)

A second concept that enters the discussion ofM I-surprises is the expectations theory of the termstructure of interest rates. This theory maintains thatsecurities of different maturities are good substi­tutes, so that competition in the financial marketswill equate the holding period yields of securities ofdifferent maturities. Thus, for example, the inves-

38

tor can expect to obtain the same yield by (I) hold­ing a six-month Treasury bill to maturity, or (2)holding a three-month T-bill to maturity and thenreinvesting the proceeds in a second three-monthbill and holding it to maturity. For this reason, theyield On the six-month bill will be equal to a weight­ed average of the yield on the three-month bill andthe expected yield on the three-month bill, threemonths from now. The latter yield is called a for­ward rate. Changes in these short-term forwardrates cause changes in the same direction in long­term rates. This term structure theory provides alink between expected future short-term rates andlong-term rates that plays an important part in thevarious explanations advanced for the effects ot'M I-surprises.

Expected Inflation EffectThe expected inflation theory holds that an Ml­

surprise is taken as new information about the cur­rent and future growth in the quantity ofMl suppliedby the Federal Reserve. As such, a positive Ml-sur­prise causes interest rates to rise because it raisesint1ation expectations. This theory has an unambig­uous implication for long-term rates and for theexchange rate. Higher expected int1ation shouldraise the former and depreciate the latter. The effectonexchange rates follows from the anticipated dropin the purchasing power of the dollar. The effects onshort-term rates would depend on how quickly theprice level can adjust to a change in money supply-that is, on how "sticky" prices are in the short­run. Presumably, if prices were sufficiently t1exible,one would observe an increase in short-term ratesfollowing a positive MI-surprise.

The empirical results contradict the expected in­t1ation hypothesis in two ways. First, the exchangerate appreciates with a positive Ml-surprise ratherthan depreciates, as predicted by theory. Second,the positive effects of Ml-surprises on long-termrates grew larger with the change in Fed operatingprocedures in 1979. IIMI-surprises changed long­term rates by altering int1ation expectations, thenthe implementation of a new anti-int1ation policyshould have reduced the estimated responses. Evenif the change in procedures had no credibility, itwould not have increased them; it would have leftthe responses unchanged.

Policy Anticipations EffectThe policy anticipations theory4 proposes that

Ml-surprises lead to changes in real interest rates.When the Fed attempts to control Ml, a positiveMl-surprise leads the market to anticipate an in­crea.se in short-term (real) interest rates as the Fedattempts to bring Ml under control. In other words,the market believes that the change in Ml wasinduced by a factor other than an intentional actionofthe Fed. Since the Fed wants to control Ml, itsubsequently will take actions to do so and its ac­tions will affect real interest rates. The anticipationof these actions causes rates to change immediately.

The maturities of the rates that are affected de­pend on what the market perceives is the source ofthe Ml-surprise. If pressures for the change in Mlare not expected to persist for long, then only veryshort-term interest rates should change. It also ispossible, for example, that the market interprets apositive Ml-surprise as an indication that GNP isstronger than it originally believed, and that thisgrowth has raised the quantity of money demanded.An upward revision in estimated current GNP couldlead to expectations of higher GNP for several yearsto come, and the market might expect the Fed tooffset these pressures through gradual and fairlyprolonged increases in short-term rates. As a conse­quence, expected forward rates would rise, andtheir rise would push up longer term spot interestrates. However, since the Fed cannot hold real inter­est rates above the equilibrium level into the verydistant future, this theory would predict that little, ifany, of the effect of Ml-surprises on forward rateswould prevail into the distant future.

Finally, the policy anticipation theory predictsthat a positive Ml-surprise causes the dollar to ap­preciate. This occurs because money surprisescause real interest rates to move, and higher realU.S. interest rates cause the demand for dollar-de­nominated assets to rise.

This hypothesis is consistent with the responsesof short-term interest rates and exchange rates. Italso is consistent with the result that both of thesevariables moved more after October 1979, when theFed's monetary control efforts were more aggres­sive, than before that date, when the Fed tended torespond more gradually. However, the strong rise oflong-term interest rates after October 1979 has beeninterpreted as contradicting this hypothesis. The

39

reasoning goes thatalthough the Fed can drive realinterest rates up by reducing the money supply inthe short-run, the effect on real rates should not beevident iri the distant future when prices have hadtime to adjust. If the period of monetary restraintlasts only a short time, then only the current interestrate .and forward rates covering the near futureshould rise; the effect on long-term interest ratesshould be small. If the Fed follows a tighter policyfar into the future, the policy should eventuallyreduce inflation expectations and lower 10hg-terminterest rates.Combined Effects

The theories discussed above are not mutuallyexclusive. One can reasonably combine the policyanticipations and expected inflation effects into anexplanation of responses of various asset prices toMI-surprises. 5 The argument is that, in the post­October 1979 period, the Fed tried to control Ml butdid so somewhat cautiously. As a result, when themarket observed a larger-than-expected increase inMl, it assumed that only part of it would be offsetby a policy response and that part of it would per­manently raise Ml. Short-term rates consequentlyrose because the Fed was expected to tighten policy,and long-term rates rose at the same time becausepart of the Ml increase was expected to remain inthe money stock permanently. Unfortunately, thiscombined theory cannot be tested with the exchangerate after October 1979 because the predictions areambiguous, that is, the increase in real rates wouldcause the dollar to apprecjate, while the increase ininflation expectations would cause a depreciation.

This explanation fits the post-October 1979 dataquite well, with both long- and short-term interestrates rising when Ml came in over expectations. Forthe pre-October 1979 period, the prediction clearlywould be for (at most) the observed small responseof the short-term rate because the Fed reacted verygradually to the MI numbers. However, the long­term rate should have responded positively, becausethe former funds rate operating procedure ofthe Fedimplied that a larger part of an Ml-surprise wouldbe permanent. By the same token, the exchange rateshould have depreciated when Ml came in overexpectations. These last two predictions of the com­bination theory do not fit the data for the pre-October1979 period-neither the long-term rate nor theexchange rate responded significantly in that periOd.

II. Forward Interest RatesOne key aspect of the puzzling reaction of finan­

cial asset prices to Ml-surprises is that short- andlong-teon rates seem to respond in the same way.Thus, prior to October 1979, neither rate respondedsignificantly, whereas afterward, they both respon­ded positively. As discussed above, the two theoriesused to explain these responses, as well as thecombination theory, do not necessarily predict thatlong- and short-rates will move in the same direc­tion. The reason is that short-term rates are taken asa measure of changes in real interest rates, whereaschanges in long-term rates are seen primarily toreflect changes in inflation expectations. The realand expected inflation components of nominal in­terest rates should respond in different ways to anMl-surprise depending upon the public's percep­tion of how policy is being conducted. Prior toOctober 1979, the Fed may have been perceived aspermitting a significant part of a positive Ml-sur­prise to remain in the money supply permanently. Inreaction, there should have been only a small re­sponse by real rates (reflected in the short-term rate)and a fairly large increase in inflation expectations(reflected in the long-term rates). After October1979, short-term rates should have respondedstrongly, whereas long-rates might have been ex­pected to respond only slightly because of the Fed'santi-inflation stance.

The assumption that the responses of short-termrates to MI-surprises reflect the real component ofinterest rates is strongly supported by evidence thatprices adjust to changes in the money supply with alag. Given this evidence, it is difficult to believethat today's money "blip" causes any perceptiblechange in the inflation expected, say, over the nextthree months.

However, the assumption that movements inlong-term rates primarily reflect changes in infla­tion expectations does not rest on such firm footing.According to the expectations theory of the termstructure of interest rates, long-term spot rates areweighted averages of the current short-term spotrate and the expected short-term forward rates.Thus, it is possible that long- and short-term ratesmove in the same direction because long-term rates,in part, are made up of short-term rates. In otherwords, changes in inflation expectations may not

40

always. dominate observed changes in long-terminterest rates. At times, observed changes may sim­ply reflect movements in short-term spot rates andfairly Ilear-term forward rates. This observationsuggests that the hypotheses concerning responsesto Ml-surprises should be tested with short-termspot rates and short-term forward rates expectedto prevail far into the future. Changes in, say,the expected one-year rate, thirty years forwardshould give an indication of movements in inflationexpectations.

Unfortunately, it is mathematically difficult tocalculate expected forward rates from the term struc­ture when one must use bonds that are couponinstruments. Forward rates have been used in anoth­er study of Ml-surprises, but the calculations weresimplified by assuming, in effect, that couponbonds were discount bonds. 6 This method providesonly very rough estimates of forward rate changes,and the result that forward rates far into the futuremoved in the same direction as long-term spot ratesshould be viewed with caution.

The study in this paper uses a method of calculat­ing forward rates from a term structure of spot rateson coupon bonds developed by Shiller, Campbelland Schoenholtz (1983).7 The authors have demon­strated that their approximation of the true formulafor calculating the desired forward rates yields closeestimates, except in cases of extreme interest ratevolatility.

Empirical ResultsIn this section, we analyze the impact of MI-an­

nouncements on the financial markets by regressionmethods similar to those that have been commonlyapplied in the literature described above, except thatwe look at the responses of short-term forward ratesrather than long-term spot rates. More specifically,we use weekly Ml-surprises as the variable to ex­plain changes on the day after the Ml-announce­ment in three financial variables-the three-monthTreasury bill rate; the two-year T-bill rate, expected5-years forward; and the ten-year T-note rate, ex­pected 20-years forward. Changes in the three­month rate are taken to reflect changes in the realcomponent. Changes in the two forward rates aretaken as reflecting changes in expected inflation.

These (ordinary least squares) regressions wererun over three sample periods, each correspondingto a different monetary policy regime. Regime Istretches from September 21, 1977 to October 3,1979, and falls in the period when the Fed used theFederal funds rate as an operating instrument. Asdiscussed earlier, this is a period when the Fedattempted monetary control only over periods ofseveral quarters through gradual movements in thefunds rate. In doing so, it tended to smooth short­run changes in the funds rate. Regime II covers

October 6, 1979 to October 3, 1982, the period ofthe nonborrowed reserves operating procedure inwhich the Fed attempted monetary control overshorter time periods, and in doing so permittedmuch more short-run variation in short-term interestrates. In Regime III, which covers October 10,1982 to February 8, 1984, nonborrowed reservesno longer were linked to Ml as they had beenbefore. This period can be considered a kind of"half-way house" between I and II. The Fed .usedborrowed reserves as its operating instrument, and

Table 1

Regression ResultsOAt a o + at UMt

IF

Regime I: 9/21/77 to 10/3/79 (98 degrees of freedom)

Three-Month Treasury Bill Rate

Two- Year Treasury Rate. Five Years Forward

Ten-Year Treasury Rate. Twenty Years FOlward

0.03 **(2.76)0.01

(0.86)

0.001(0.012)

5.79**(2.59)025

(0.16)

1.34(050)

0.05

CU)I

0.01

Regime II: 10110/79 to 10/3/82 ( 135 degrees offreedom)

Three-Month 1reasury Bill Rate

Two- Year Treasury Rate. Five Years Forward

Ten-Year Treasury Rate. Twenty Years Forward

0.04(I (8)

0.03(150)

0.Q2(0.36)

36.45**(638)

7.48*(2.08)

2.21(0.21)

0.23

0.02

0.01

Regime III: 10/10/82 to 2/8/84 (52 degrees of freedom)

Three-Month Treasury Bill Rate

Two-Year Treasury Rate, Five Years Forward

Ten-Year Treasury Rate, Twenty Years Forward

0.ClO4(0.35)

0.02(0.73)

0.01(0.21 )

16.31**(5.45)

12.57*(2 17)

7.17(0.99)

0.34

0.06

0.001

UM weekly pereentage change in M I minus the expected weekly percentage change (latter variable defined as the medianforecast of the survey of money market economists conducted by Money Market Services, Inc.).

DA change in specified variablewhere, the changes are calculated for Friday over Thursday in weeks prior to 1/30/80, and for Monday over Friday inweeks ending 1/30/80 and after. When the day of or the day following an MI announcement fell on a holiday, data for thatweek were excluded from the sample.

Note: * significant at the 5-percent level* * significant at the I-percent level

statistics in parentheses.

41

beyond seven years are not statistically significant,indicating that a money surprise today has no sys­tematic effect on inflation expectations beyond sev­en years. The point is illustrated by the result pre­sented in Table 1, that the lO-year rate, twenty yearsforward, has a coefficient ofonly 2 basis points withat-statistic of only 0.21. (Other results not present­ed show that the three-year rate, seven years for­ward,a.lso was not significantly influenced by MI­surprises).

expected, the results for Regime III show asmaller response for the real component (as reflec­ted)nthe 3-month T-bill rate) and a larger responsein inflation expectations (as reflected in the forwardrates).

Specifically, a one-percent positive MI-surprisecauses the three-month T-bill rate to rise by 16 basispoints (less than in Regime II) and the two-yearrate, five years forward, to rise by 12Vl percent(more than in Regime II). As with Regime 11, thereis no significant response of the ten-year rate,twenty years forward.

However, the real puzzle occurs in Regime I, thefederal funds rate period. The apparent perceptionby the market that the Fed reacted cautiously toM I-surprises did not translate into changes in infla­tion expectations when M I came in over or underexpectations. In that period, a one-percent MI-sur­prise caused a small, but statistically significant,6-basis-point increase in the Treasury bill rate, andno statistically significant change in the two for­ward rates reported.

permitted more short-run variability in interest ratesthan in Regime I, but less than in II. 8

Thus, purely on the basis of considering monetarypolicy regimes, we would expect to find the largestincreases inshort-term rates in response to positiveMl-surprises to occur in Regime II, the next largestin Regime III and the smallest in Regime I. The sizeof these hypothesized responses reflects the hypo­thesized market perception ofthe Fed's commit­menno controlling money in the short-run, that is,the greatest perceived commitment is demonstratedby the greatest willingness to permit short-terminterest rates to move in response to a money"blip." Moreover, we might expect that forwardrates would respond most strongly in the funds rateRegime I, less strongly in Regime III and the leaststrongly in Regime II.

The empirical results are presented in Table 1.These results seem to resolve some, but not all, ofthe money supply announcement puzzle. The re­sults for Regimes II and III seem to make sense interms of the theoretical expectations discussed ear­lier. In the nonborrowed reserves Regime II, a onepercent positive Ml-surprise caused a substantial36-basis-point increase in MI on the following day.Moreover, this response is highly statistically signi­ficant (the t-statistic is 6.38).

This large response in the real rate contrasts withthe far smaller increase in expected inflation, asmeasured by the 7Vl basis point increase in thetwo-year rate, five years forward. Although thisresponse is fairly small, it is statistically significant(the t-statistic is 2.08). Forward rates extending

III. Large Structural Deficits andthe Effects of Money Supply Announcements

The preceding analysis of the impact of MI-an­nouncements on asset prices separated the sampleperiod at October 1979 and October 1982 on thetheory that the Federal Reserve has a significantinfluence on those prices. With different operatingprocedures, the market presumably anticipates dif­ferent Fed behavior and asset prices respond dif­ferently. This is quite a reasonable presumption.However, the Fed's change in operating procedureswas not the only major policy change affecting thedata in the 1980s. There also was a major change infiscal policy, as reflected in the emergence of largesustained federal deficits. Expectations of future

42

large structural deficits were well-formed at least asearly as President Reagan's tax cuts in July 1981.

How could the presence of large structural defi­cits affect the response of asset prices to moneysupply announcements? They could affect the re­sponse of asset prices if the public believes theFederal Reserve may monetize part of the federaldebt that is generated by budget deficits. Someevidence of past monetization is in the economicsliterature, although it is by no means conclusive,and expectations concerning monetization are fre­quently voiced in the financial press. 9 For example,Hoey and Hotchkiss (1983) report results of a sur-

vey of financial decision-makers in December1983, which found that out ofover600 respondents,about two-thirds agreed with the following state­ment "One can have no confidence in the stayingpowerof disinflationary monetary policy as long asfederal deficits remain in triple digits."

Monetization of the debt obviously would resultin higher inflation rates in the long-run, and thushigher long-term interest rates. Once it becomesclear that deficits are and will remain large, bondrates should rise if the public believes the monetiza­tion hypothesis. What role do MI-announcementsplay? Although the public might believe the moneti­zation hypothesis as ageneral principle, it presum­ably would be uncertain about how much of thefederal debt might be monetized. Such uncertaintywould be consistent with the apparently changingrelationship between deficits and MI growth in thepast. Moreover, changes in the make-up of theFederal Open Market Committee, and in the generalpolitical "climate" may be expected to influencethe degree of monetization.

As the weekly stock of MI is announced, thepublic may refine its views about the degree ofmonetization, and therefore about future moneygrowth and inflation. Thus, if MI comes in higher(lower) than expected, the market may revise up(down) its estimate of the degree of monetization,and as a consequence, long-term rates will rise(fall). The effects on long-term rates would not beobserved in a period of low budget deficits becausethere would not be much pressure for monetizationin such a period. With respect to short-term rates, apositive M I-surprise might elicit smaller responseswhen deficits are large because the market expects aless aggressive offsetting action by the Fed.

As pointed out by Hardonvelis, 1982, large struc­tural deficits may help to make more sense out of thecombined inflation expectations/policy anticipationexplanation of asset price movements. In the 1980­1982 period, a positive MI-surprise may havecaused short-term rates to rise because of anticipa­tions ofa partially offsetting action by the Fedunder the reserve control procedures. The samepositive MI-surprise may have caused forward ratesto rise because the public revised up its estimate ofhow much of the government debt would be mone­tized, and thus its view of inflation in the long-run.In the pre-1980 period, short rates may have

43

responded only slightly because no immediateFed policy reaction was anticipated, and forwardrates may have responded only slightly becausedeficits were small and there was little pressure formonetization.

A Test of the Monetization HypothesisSimply stated, the monetization hypothesis is

that a positive (negative) MI-surprise will elicit asmaller increase (decrease) in short-term rates and alarger increase (decrease) in forward rates whendeficits are "large" than when they are small. Thishypothesis can be tested using data from monetarypolicy Regime II. If we assume that the dividingline between "small" and "large" expected futuredeficits was designed by the Reagan tax cuts of July1981, then we can test the monetization hypothesisby using data from monetary policy Regime II­October 1979 to October 1982. Specifically, we cansee whether the estimated responses of spot andforward interest rate and exchange rates changed aspredicted in mid-198I. To do so, we estimate thesame model as in Table lover the period October10, 1979 to October 3, 1982, with two additionalarguments: (I) a dummy variable (D), that is zeroprior to July 6, 1981 and llnity thereafter; and (2)the MI-surprise variable (UM) multiplied by thesame dummy variable (D). The first of these twoadditional arguments permits· the intercept teill1 toshift, while the second additional argument permitsthe estimated response of interest rates to MI-sur­prises to shift. The t-statistics on these coefficientswill indicate whether the estimated shifts are statis­tically significant.

The regression results are presented in Table 2.The results for the short-term spot interest rate areconsistent with the.. monetization hypothesis. Priorto the Reagan tax cut in mid-1981, a one-percentMl-surprise induced a 44-basis-point increase inthe three-month Treasury bill rate. After the tax cut,this policy anticipations effect is estimated to bemuch smaller-the effect falls by 22 basis points.The same basic result is obtained for the one-yearrate, one-year forward: before the tax cut the re­sponse was 32 basis points, and after the tax cut itfell by a statistically significant 22 basis points.

The results for the more distant forward rates,however, seem to be inconsistent with the moneti­zation hypothesis. If,in the presence oflarge defi-

DA

Table 2

Regression Results

OAt = a o + a l UMt + f300t + f310t UMt

f30 f31

Sample Period: 10/10179 to 10/3182 (135 degrees of freedom)

Three-Month 0.06 -0.05 44.21 ** -22.45* 0.24Treasury Bill Rate (1.55) (0.86) (6.38) (1.87)

One-Year Treasury -0.003 0.()O4 32.43** -21.69** 0.28Rate, One-Year Forward (0.13) (0.009) (7.25) (2.82)

Two-Year Treasury 0.017 0.034 9.05** -5.03 0.02Rate. Five Years Forward (0.65) (0.87) (2.03) (0.66)

Ten-Year Treasury 0.004 0.07 10.90 -9.36 0.01Rate, Twenty Years Forward (0.06) (064) (0.94) (099)

D, = 0 in 10/10/79 to 6/30/81, and I in 7/6/81 to 10/3/82.

Note: Variables are defined in note to Table I.

* significant at the 5-percent level (one-tailed test)** significant at the I-percent level (one-tailed test)

cits, the public interprets a positive Ml-surprise asraising the odds that the Fed is monetizing part ofthe deficit, then it simultaneously should raise itsexpectations of future inflation. For this reason,coefficient f3 I in Table 2 for the two-year rate, fiveyears forward (and possibly for the ten-year rate,twenty years forward) should be significantly posi­tive. Instead, it is insignificantly negative-that is,there was no statistically significant shift in theresponses of this variable corresponding to the taxcut in mid-I98l.

On the basis of the evidence presented, it does notappear that the presence of large structural deficitsresolves the inconsistencies that appear to exist inthe responses of forward interest rates to Ml-sur-

prises. The main mystery that still exists is thefollowing: if a positive Ml-surprise after the Fedbegan exercising better long-run control over Ml inOctober 1979 caused an increase in inflation expec­tations, why did it not also do so prior to that date,when the Fed's control procedures seemed less welldesigned to control inflation? The monetization hy­pothesis holds that some of this inconsistency couldbe resolved by the presence of large deficits and thefear that they might be monetized after mid-1981,and the lack of large deficits in the earlier period.However, there does not appear to have been achange in the response of inflation expectations toMl-surprises between the two periods.

1\1. ConclusionThis paper has examined the responses of interest

rates to announcements of changes in M 1 that arenot anticipated by the market. The findings can besummarized as follows. Short-term spot rates ofinterest increase in response to a positive M I-sur­prise, and like earlier studies, we found that theresponses became much larger when the FederalReserve used a nonborrowed reserves-oriented op-

44

erating procedure in October 1979 to October 1982than in the earlier Federal funds rate regime. TheFed's current operating procedure, which is ori­entedaround borrowed reserves, seems to be inter­preted by the market as a kind of "half-way house"between the previous two regimes. The responses ofshort-term spot rates are larger than in the funds rateregime, but smaller than in the nonborrowed re-

serves regime. These positive responses of short­t$rm rates,together with their relative sizes in thevarious monetary policy regimes, strongly suggestthat the reSponses represent a policy anticipationseffect. That is, when Mt comes inoverexpecta­tions, the market expectsthe Fed to tighten policy tosome extent, depending on the policyregime.

The interpretation of the responsesof long-termrates is not as straightforward. Earlier studies havefound a highly significant, .and surprisingly large,positive· response of long-term interest rates, evenout to maturities of thirty years, in response to M IsUrprises in the post-October 1979 period, but littleeffect before. These responses sometimes havebeen interpreted as demonstrating changes in infla­tion expectations. It is difficult to imagine why asingle weekly M 1 figure would have a substantialimpact on long-run inflation expectations. To castmore light on this issue, we examined the responsesof forward rates of interest far in the future. It doesappear that inflation expectations were affected byM I-surprises after October 1979, but these effectsseem to extend out only about seven years, and to beof a fairly reasonable size. Thus, the responses of30-year bonds does not suggest that expected infla­tion thirty years hence has changed. Instead, theresponses reflect changes in realinterest rates cur­rently and in the near future, and inflation expecta­tions out to about seven years.

This combination of responses is consistent withthe view that the market believes the Fed has at­tempted to control Ml somewhat cautiously. A pos­itive Ml-surprise apparently causes the market toexpect some tightening action by the Fed, but notenough tightening action to prevent a moderate in-

45

crease in inflation. This interpretation is confirmedby the result that when the Fed switched from anonborrowed reserves- to a borrowed reserVes-ori­ented operating procedure, the policy anticipationeffectbecame smaller arid the inflation expectationseffect larger. The market's apparent perception thatthere would be a less aggressive tighteningofpolicywhen M I increased unexpectedly therefore corres­ponded to the anticipation that infl:j.tion wouldin.­crease by more than itwould have in the earlierpolicy regime.

The remaining puzzle about Ml-announcementeffects is that there appears to have been no re­sponse in inflation expectations prior to October1979, when the Fed used a funds rate operatingprocedure. Since that procedure involved. only· avery small policy anticipations effect, one mightexpect a large inflation expectation effect, especi­ally compared with the period after October 1979when the Fed explicitly pursued an anti-inflationpolicy.

This paper hypothesized that the lack of a stronginflation expectation effect may be related to thechange in fiscal policy regimes in mid-1981, whenexpected future structural budget deficits clearlybecame "large." It is possible that the public fearsmonetization of the government debt when deficitsare large, and that a positive MI-surprise tends toadd to this fear. Conversely, the lack of large defi­cits in the funds rate regime might help explain thelack of an inflation expectation effect. The resultsfor short-term spot rates seem to confirm this hypo­thesis, but those for expected future interest rates donot. Further research will be required to solve themoney supply announcement puzzle completely.

FOOTNOTES

1. See Cornell, 1983, for a review of this literature.

2. See DeRosa and Stern, 1977.

3. See Judd, 1982.

4. Cornell, 1983, discusses two other theories. The realactivity hypothesis argues that the demand for money is afunction of expected future income, and that an M1-surprisecauses an upward revision in estimates of expected futureincome. As a consequence, interest rates rise when M1rises more than expected. Cornell also discusses the riskaversion hypothesis, which argues that an unanticipatedincrease in M1 reveals that aggregate risk aversion hasincreased.

5. See Hardouvelis, 1982.

6. Ibid

7. For example, the ten-year rate, twenty-years forward iscalculated as a duration weighted average of the differencebetween the thirty-year spot rate and the twenty-year spotrate. The linear approximation enters the formula for calcu­lating duration. This method also is employed in Loeys,1984, which came to the present author's attention as thispaper was going to the printer.

8. See Wallich, 1984.

9. See Hamburger and Zwick, 1981, McMillian and Beard,1982, and Niskanen, 1978.

46

REFERENCES

Bradford Cornell, "The Money Supply Announcement Puz­zle: Review and Interpretation," American EconomicReview, September 1983, pp. 644-57.

Paul DeRosa and Gary Stem, "Monetary Control andthe Federal Funds Rate," Journal of MonetaryEconomics, April 1977, pp. 217-30.

Gikas Hardouvelis, "Interest Rate Volatility, Fed Credibility,and the Weekly Monetary Announcements," unpub­lished manuscript, University of California, Berkeley,1982.

Michael J. Hamburger and Burton Zwick, "Deficits, Moneyand Inflation," Journal of Monetary Economics, Jan­uary 1981, pp. 141-50.

Richard B. Hoey and Helen Hotchkiss, "Decision-MakersPoll," A. G. Becker Paribas, Inc., December 21, 1983,pp.3-5.

John P. Judd, "An Examination of the Federal Reserve'sStrategy for Controlling the Monetary Aggregates,"Economic Review, Federal Reserve Bank of SanFrancisco, Fall 1982, pp. 7-18.

Jan G. Loeys, "Market Perceptions of Monetary Policy andthe Weekly M1 Announcements," Working Paper No.84-2, Federal Reserve Bank of Philadelphia.

W. Douglas McMillin and Thomas R. Beard, "Deficits,Money and Inflation Comment," Journal of MonetaryEconomics, 1982,10, pp. 273-277.

William A. Niskanen, "Deficits, Government Spending, andInflation," Journal of Monetary Economics, 1973,4,pp.591-602.

Henry Wallich, " Recent Techniques of Monetary Policy,"Economic Review, Federal Reserve Bank of KansasCity, May 1984, pp. 21-30.