Short–Run Effects of Fiscal Policy with Forward–Looking ...€¦ · Short–Run Effects of Fiscal Policy with Forward–Looking Financial Markets 357 National Tax Journal Vol.

Short–Run Effects of Fiscal Policy with Forward–Looking Financial Markets

357

National Tax JournalVol. LV, No. 3September 2002

Abstract - We explore the short–run effects of fiscal policy usingsimulations of an empirical, rational–expectations, open–economymacromodel developed at the Federal Reserve Board. Based on thismodel, we find that tax cuts and spending increases generally stimu-late economic activity in the short run, contrary to the extremeview that forward–looking financial markets more than offset thedirect expansionary impulse of those actions. However, the magni-tude of the stimulus is greatly attenuated by the financial–marketfeedback. For example, a sustained cut in personal income taxesraises output by less than the amount of the tax cut itself, and itlikely reduces output (relative to baseline) in the first several yearsif phased in gradually over time. Our results also show that theestimated stimulus imparted by fiscal policy is sensitive to reason-able variation in the model’s parameters.

INTRODUCTION

Traditional analyses of fiscal policy imply that raising gov–ernment spending or reducing taxes stimulates economic

activity in the short run. However, an alternative view de-veloped by Blanchard (1984) and Branson (1985) emphasizesthat such policies also induce responses in capital marketsthat tend to diminish economic activity. The idea is straight-forward: An expectation of larger future budget deficits boostsfuture short–term interest rates, and these higher future short–term rates boost current long–term interest rates. Higher long–term rates damp business investment and other interest–sen-sitive spending, offsetting at least part of the direct expan-sionary effect of the tax cut or spending increase.

This alternative view raises the possibility that purport-edly expansionary fiscal policies could actually reduce eco-nomic activity in the short run, depending on the time pathof the fiscal actions and the degree to which financial mar-kets respond to changes in expected long–run conditions. Inthe extreme, it might be possible that the traditional recom-mendation for countercyclical fiscal policy has the oppositeof its intended effect, with economic stimulus coming fromsmaller budget deficits rather than larger ones. Theory alonecannot predict whether the indirect damping effects causedby forward–looking financial markets will more than offsetthe direct stimulative effects of higher government spendingor lower taxes. However, theory suggests that this outcome

Short–Run Effects of Fiscal Policy withForward–Looking Financial Markets

Douglas W.Elmendorf &David L.ReifschneiderFederal Reserve Board

NATIONAL TAX JOURNAL

358

is more likely if the fiscal action is phasedin over time, because the direct stimulustakes hold only gradually, while the de-pressing effect of the financial–market re-action occurs immediately.

Economists in several presidential ad-ministrations have invoked this logic toargue that seemingly contractionary fis-cal policies might spur economic activityin the short run. Nearly twenty years ago,the 1984 Economic Report of the President(pages 40–41) stated that a crediblephased–in deficit reduction plan wouldnot hamper near–term economic growth.Almost ten years ago, the 1994 EconomicReport of the President (pages 34–38) statedthat the deficit reduction package enactedin the preceding year had brought downlong–term interest rates and that the “de-clines in long–term interest rates that haveoccurred since the 1992 election . . . aremore than enough to offset the con-tractionary effects of this decrease in thestructural deficit.”1

A number of recent papers have ad-duced new empirical evidence on the netstimulative effect of various fiscal actions.For example, Ramey and Shapiro (1998),Edelberg, Eichenbaum, and Fisher (1998),Fatas and Mihov (1999), Blanchard andPerotti (2001), and Mountford and Uhlig(2002) use vector autoregressions in dif-ferent ways to study the macroeconomiceffects of increases in government spend-ing or decreases in taxes. However, thetechniques used in these papers cannotdistinguish effectively among tax cuts orspending increases with different timepatterns or different compositions, in partbecause they do not control explicitly forexpectational effects related to future fis-cal policy actions and other factors. An-other line of research—including Alesina

and Perotti (1997), Giavazzi, Jappelli, andPagano (2000), and others—has examinedcase studies of fiscal retrenchments andexpansions. Hemming, Kell, and Mahfouz(2000) review the literature on the effec-tiveness of fiscal policy in stimulating eco-nomic activity and conclude that: “Theempirical evidence from advanced econo-mies suggests that fiscal multipliers aretypically positive but small . . . While thereis some evidence of negative fiscal multi-pliers, . . ., the preconditions for such anoutcome cannot be precisely pinneddown” (page 3).

In this paper we explore the effectsof fiscal policy over a five–year periodusing simulations of an empirical, ratio-nal–expectations, open–economy, struc-tural macromodel developed at theFederal Reserve Board. We examine avariety of fiscal actions: a sustained in-come–tax cut, a one–time income tax re-bate, an investment tax credit, and asustained increase in government pur-chases.2 Our findings are unavoidablysensitive to the specification of the model,so we present results under different as-sumptions about the sensitivity of con-sumption to short–run fluctuations in in-come, the sensitivity of international capi-tal flows to shifts in U.S. interest rates, andother factors. Our results are also condi-tional on the monetary policy that accom-panies the fiscal actions. We assume thatthe federal funds rate is set according to asimple policy rule under which a fiscalaction that successfully stimulates theeconomy will induce only a modest risein short–term interest rates. Experimen-tation with monetary policy rules that areeven less responsive in the short runyields essentially the same qualitativeconclusions.

1 Elmendorf, Liebman, and Wilcox (2002, pgs. 72–77) review the role played by this alternative view of fiscalpolicy during early discussions of deficit reduction and economic stimulus in the Clinton administration.Summers (2000) provides the administration’s retrospective view of the effects of fiscal policy in the 1990s.

2 We use the term “sustained” to describe tax cuts or spending increases that last more than ten years. We do notexamine fiscal changes that are truly permanent, because these actions—if not accompanied by offsettingbudget adjustments—yield long–run movements in the debt–GDP ratio that are extremely large.


359

Our conclusions can be summarizedbriefly. First, in our model a sustained cutin personal income taxes provides only asmall economic stimulus in the short run,and it reduces output in the longer run rela-tive to baseline. The estimated short–runtax–cut multiplier is well below one (inabsolute value) in most of our simulations,suggesting that variations in taxes largeenough to generate substantial short–runmacroeconomic effects would have largebudget consequences, and thus could sig-nificantly impede other objectives of taxpolicy such as efficiency and equity. Sec-ond, forward–looking financial marketsoffset much of the incipient short–runstimulus of a sustained personal income taxcut—roughly half in our example. Indeed,such a tax cut is mildly contractionary ifphased in over an extended period, becausethe indirect damping effect through finan-cial markets slightly precedes the directstimulus from higher income. This resultimplies that some back–loaded fiscal ex-pansions may be contractionary, and someback–loaded contractions may be expan-sionary. Third, a one–time income tax re-bate creates a significant initial stimulus ifa majority of consumption tracks currentincome, but very little stimulus if it doesnot. Fourth, a sustained tax incentive forbusiness investment has little effect on eco-nomic activity for about a year but gener-ates substantial stimulus thereafter. Fifth,a sustained increase in government pur-chases boosts output immediately when theadditional outlays occur; the practical im-port of this finding is tempered by the sub-stantial lags in deciding on changes inspending and then disbursing the funds.3

Again, these results are sensitive to thespecification of the model used in the analy-sis; alternative specifications might yielddifferent conclusions.

The next section of the paper brieflydescribes the relevant characteristics of themodel we use, and the following section

discusses the relationship between bud-get deficits and interest rates in the model.Two ensuing sections present our simu-lation results, and a final section con-cludes.

A FRAMEWORK FOR ANALYSIS

FRB/US is a large–scale, quarterly,econometric model of the U.S. economy,developed at the Federal Reserve Board.The model was designed to accord witheconomic theory and to fit the data well—both equation–by–equation and for theentire system of equations. As discussedby Brayton, Levin, Tryon, and Williams(1997), the autocorrelations and momentsgenerated by the model line up reasonablywell with historical observations, and thedynamics of the model as a whole conformclosely with the dynamics of reduced–formvector autoregressions. This combinationof empirical fit and theoretical structuremakes the FRB/US model an especiallyuseful tool for evaluating the effects of al-ternative fiscal (and monetary) policies.

For our analysis, a critical feature of themodel is that households and firms are for-ward– looking, and base their decisions onexpectations of future income, sales, finan-cial conditions, and so on. The simulationsin this paper assume that expectations areformed rationally—that is, in a mannerconsistent with the structure of the model.However, the presence of adjustment costsimplies that households and firms respondonly gradually to changes in expectationsand other economic factors. As a result, theeconomy is New Keynesian in the shortrun (with labor and product markets notalways in steady state, and inflation inher-ently persistent) and neoclassical in thelong run (with markets returning tosteady–state equilibrium over time). FRB/US is not strictly a dynamic, stochastic,general–equilibrium model, but it is withinhailing distance of one.

3 See Taylor (2000) for other concerns about discretionary countercyclical fiscal policy.


360

In this paper we simulate the FRB/USmodel under alternative fiscal policies andcompare the outcomes to a baseline inwhich the economy is in long–run equi-librium. Because the model is essentiallylinear, the effect of fiscal policy on eco-nomic conditions is mostly independentof the economy’s starting point, so the re-sults we report also reflect the impact offiscal policy during recessions andbooms.4 In the remainder of this sectionwe summarize briefly the aspects of FRB/US that are most relevant to the subject ofthis paper. General descriptions of themodel can be found in Brayton andTinsley (1996), Brayton, Levin, Tryon, andWilliams (1997), Brayton, Mauskopf,Reifschneider, Tinsley, and Williams(1997), and Reifschneider, Tetlow, andWilliams (1999); additional information isavailable upon request.5

Household Spending

Consumer spending on nondurablegoods and services in FRB/US follows anerror–correction framework in which ac-tual consumption adjusts slowly towardtarget consumption. Because householdsare forward–looking, their target con-sumption depends on perceived perma-nent income and property wealth. Perma-nent income includes separate terms forlabor income, transfer income, and prop-erty income.6 The formulas for permanentincome discount future income 25 percentper year, a value chosen through a grid

search to maximize the fit of the consump-tion equation. This discount rate is muchhigher than one would find in a standardperfect–foresight lifecycle model, but it isappropriate if households are highly riskaverse and face uninsurable income un-certainty. Indeed, Carroll and Summers(1991) and Carroll (2001) note thatFriedman’s (1957, 1963) original statementof the permanent income hypothesis useda high discount rate, effectively settingpermanent income equal to average ex-pected income over the following half–dozen years.

Owing to various frictions (perhapshabit persistence in part, although this isnot modeled explicitly), actual consumerspending adjusts only half of the way to-ward its target level after one year. Spend-ing also responds to the expected cyclicalposition of the economy: Higher expectedunemployment over the next few yearsholds down spending today, perhaps be-cause it signals a heightened uncertaintyabout each household’s future income andthus raises the discount rate applied to it.Lastly, about ten percent of spending isdone by households whose consumptionequals their income. The overall short–term sensitivity of consumer spending tocurrent income in FRB/US is somewhatlarger than ten cents on the dollar, becauseof the cyclical effect just described andbecause current income affects permanentincome. Still, the estimated response ismuch smaller than Campbell andMankiw’s (1989) finding that about half

4 To be clear, the cyclical position of the economy affects households’ and firms’ behavior, but it does not muchaffect their reaction to changes in fiscal policy.

5 It is worth stressing that the results reported in this paper are based on the version of FRB/US in place in May2002. Previous versions of FRB/US would have given somewhat different quantitative answers to the ques-tions posed in this paper; no doubt future versions will as well. These changes in model properties over timeare a natural consequence of continued economic research and model development. For example, the originalversion of FRB/US introduced in 1995 had a simpler supply–side structure than that in place today. Thischange, coupled with modifications to the international sector and to other equations of the model, has mod-erated the effect of fiscal policy on long–run real interest rates.

6 Because property wealth enters separately, the inclusion of permanent property income might seem redun-dant. However, its estimated coefficient is economically and statistically significant. Moreover, because prop-erty income equals the after–tax rate of return multiplied by property wealth, its inclusion can be viewed asan indirect way to make consumption depend on the return to saving.


361

of consumption is done by householdswho follow the “rule of thumb” of settingconsumption equal to income. Someof our simulations raise the effectiverule–of–thumb share to 50 percent, mak-ing the model’s consumption dynamicssimilar to those of the Campbell–Mankiwmodel.

Household outlays for consumer du-rable goods and housing are derived byassuming that the desired stocks ofdurables and housing are proportional tooverall target consumption, with a pro-portionality factor that depends on theuser cost of those items. User cost incor-porates interest rates, relative prices, and(for housing) the tax deductibility of mort-gage interest payments. Outlays displayaccelerator–type behavior, in whichspending overshoots its long–run targetin order to achieve a more rapid adjust-ment of the stocks of durables and hous-ing.

Consumer spending on nondurablegoods and services does not respond di-rectly to interest rates or the personal taxrate on capital income. However, shifts ininterest rates matter for spending indi-rectly through their effect on wealth andexpected property income (factors thathelp determine the long–run target levelof spending). Similarly, changes in capi-tal income taxes matter for spending in-directly through their effect on expectedafter–tax property income. Through thesechannels, increases in the after–tax rate ofreturn on saving raise the aggregate pri-vate saving rate, all else equal.

Business Investment

Businesses in FRB/US determine theirinvestment by maximizing expected

profits subject to costly adjustment of thecapital stock. The target ratio of capitalto output depends on the parameters ofthe Cobb–Douglas production functionand on user cost.7 The user cost incorpo-rates the cost of financing, the price ofcapital, the depreciation rate, and the cor-porate tax system; it does not include per-sonal income taxes, so changes in thosetaxes in the model do not affect invest-ment directly. Firms’ financing cost ismeasured as a weighted average of thecosts in debt and equity markets; a gridsearch implies that roughly equal weightsoptimize the fit of the investment equa-tion. The real cost of debt finance is thetax–adjusted nominal yield on BAA cor-porate securities less expected inflation;the real cost of equity finance is the nomi-nal yield on Treasury bonds plus an esti-mated equity premium, less expected in-flation.

Because of planning lags, engineeringrequirements, and other costs of adjust-ment, business investment depends onboth past and expected future valuesof the target capital–output ratio. Theestimated equation implies that aboutone–sixth of the adjustment of equipmentcapital to a shock to output or user cost iscompleted within one year; adjustment ofnonresidential structures is considerablyslower.

Government Sector

The major categories of governmentspending and taxes vary cyclically inFRB/US around their long–run trends.Transfer payments and tax collections actas automatic stabilizers, while spendingon goods and services is estimated to beslightly procyclical.8

7 The estimated business investment equations are derived from a production function that has a Cobb–Douglasstructure for labor, energy, and aggregate capital services. The capital services aggregate is a function of fourtypes of capital using a translog approximation that allows for substitution based on relative prices. Becausethis substitution is driven primarily by secular declines in the relative price of computers, it plays little role inthe analysis of fiscal policy, and we simplified the capital equations to Cobb–Douglas for the current paper.

8 See Cohen and Follette (2000) for a discussion of automatic fiscal stabilizers.


362

Aggregate Supply, Wages, and Prices

Potential output in FRB/US depends onactual capital services, potential labor in-put, potential energy input, and trendmultifactor productivity. According toFRB/US, trend multifactor productivity isnot affected by the fiscal policies we ex-amine, and we also do not consider anychanges in potential energy input. How-ever, our results do incorporate the effectsof fiscal policy on capital services—through changes in housing and businessinvestment—and on potential labor in-put—through tax–induced changes inhours worked.

For the version of FRB/US used in thispaper, we set the long–run uncompen-sated elasticity of trend hours workedwith respect to (one minus) the marginaltax rate at 0.17, as in Gale and Potter(2002).9 These authors briefly review thevery large literature on wages and laborsupply, including research on the effect ofchanges in tax rates. In a nutshell, this lit-erature generally finds a small elasticityof hours worked for men, a somewhatlarger elasticity for women working full–time, and an even larger elasticity forwomen not working full–time who canmove along the participation margin aswell as the hours margin. Gale andPotter’s aggregate estimate of 0.17 is theweighted average of their chosen elastici-ties of 0.05 for men, 0.3 for women whowork full–time, and 0.8 for other women.10

We assume that households adjust poten-tial labor supply in the same sluggishmanner as consumption, so that only half

of the distance from the target is closedwithin one year.

In FRB/US, actual labor input in pro-duction depends on firms’ decisions abouthow much labor to employ. Labor inputresponds only gradually to movements inpotential labor supply and—in keepingwith the empirical evidence on laborhoarding during business cycles—tomovements in output. In other words, theestimated labor demand and supply equa-tions in the model are consistent withOkun’s law.

In equilibrium, workers are paid thevalue of their marginal product, andprices are set according to a markup overtrend unit labor costs. However, labor con-tracts and other rigidities slow the speedat which wages and prices adjust to shiftsin supply and demand. As a result, cycli-cal dynamics in FRB/US are consistentwith the New Keynesian approach tobusiness cycles.

Monetary Policy and Financial Markets

Our simulations assume that the Fed-eral Reserve sets the federal funds rate inaccordance with a simple policy rule ofthe sort analyzed by Taylor (1993) and oth-ers. In particular, we assume that thefunds rate is raised above its equilibriumvalue (defined as the policymaker’s esti-mate of the equilibrium real rate plus thecurrent rate of inflation) by half of thedifference between the actual and targetinflation rates, and by half of the percent-age difference between actual and poten-tial output.11 This coefficient on the out-

9 By contrast, in the standard version of FRB/US the wage elasticity of hours supplied is assumed to be zero.This restriction ensures that, in the long run, labor force participation does not rise unchecked in the face ofongoing technological progress that causes the real wage to grow continually over time.

10 Auerbach and Feenberg’s (2000) analysis of federal taxes as automatic stabilizers assumes that the elasticity oflabor supply with respect to temporary changes in tax rates is between 0.3 and 1.0, because the income effect issmall from a lifetime perspective and because people can substitute their labor supply intertemporally. Ouranalysis focuses on sustained changes in tax rates to which these arguments do not apply.

11 The policymaker’s estimated equilibrium real interest rate used in our version of the Taylor rule is not fixed,but adjusts gradually over time toward the true equilibrium real interest rate—that is, the value of the realfunds rate consistent with stable inflation and the unemployment rate equal to the NAIRU. This graduallearning rule ensures that macroeconomic stability is achieved in the long run, but it prevents the funds ratefrom jumping upon enactment of fiscal policy that alters the economy’s true equilibrium real rate.


363

put gap is the one proposed originally byTaylor, but it appears to understate theresponsiveness of monetary policy to theoutput gap during the past decade, whichis more consistent with a coefficient ofunity. We view the relative restraint of thisassumed reaction function as a virtue forthis paper, because it allows the short–runeffects of alternative fiscal policies to showthrough more clearly in our results. At thesame time, assuming that monetary policyis completely unresponsive to economicconditions would not be reasonable, be-cause the economy would become un-stable in the long run. One of our simula-tions illustrates the effect of a more de-layed monetary response.

The markets for bonds, equities, and for-eign exchange are forward–looking inFRB/US. As a result, anticipated future val-ues of the federal funds rate influence thecurrent prices and yields on financial assetsthrough standard arbitrage relationships.

Long–term interest rates adhere to theexpectations theory of the term structure.Yields on Treasury securities equal aweighted average of the expected futurefunds rate, adjusted for time–varying termpremiums that move inversely with theexpected cyclical state of the economy.Yields on corporate securities are con-structed in a parallel fashion, and rates onhome mortgages key off the 10–year Trea-sury yield.

Equity prices equal the present dis-counted value of future dividend pay-ments, where the discount rate equals the

corporate bond rate plus an equity pre-mium. In the simulations, the equity pre-mium is assumed to be fixed at approxi-mately 2 1/2 percent.12 Firms have a tar-get dividend–payout ratio and smoothactual dividends through transitory fluc-tuations in earnings.

The short–run path of the real exchangerate is determined by the condition foruncovered interest parity. This conditionimplies that, apart from risk, if the ex-pected yield on a dollar–denominatedbond is higher than the yield on a bonddenominated in a foreign currency, inves-tors will hold the foreign bond only if thedollar is expected to depreciate enough toequate the two yields when measured ina common currency.13 The long–run valueof the dollar is determined by purchasingpower parity. Therefore, when domesticinterest rates rise relative to foreign rates,the dollar immediately appreciates to en-sure the requisite amount of future depre-ciation and the same long–run exchangerate. FRB/US assumes that the durationof the average bond is six years, so a 1percentage point increase in the spreadbetween domestic and foreign long–terminterest rates causes a 6 percent apprecia-tion of the dollar, all else equal. The valueof the dollar is also influenced by the ra-tio of net foreign assets to GDP, acting asa proxy for country risk. In the long run,this relationship implies that domestic realinterest rates equal foreign real interestrates plus a premium, where the magni-tude of the premium depends on U.S. for-

12 The FRB/US formulas for the cost of capital and the valuation of equities are based on the traditional view ofdividend taxation, in which such taxation represents a marginal burden on business investment. Under thisview, a reduction in the personal income tax rate has no effect on equity valuation but pushes down the costof capital (see Poterba and Summers, 1985). An alternative, so–called “new,” view of dividend taxation assertsthat dividend taxes are capitalized into the value of equities and do not affect the marginal incentive to invest.Under this view, a reduction in the personal income tax rate raises equity values but affects the incentive toinvest only in the short run (see Auerbach, 1979, Bradford, 1981, and King, 1977). Analysts remain divided onwhich of these perspectives—or some other perspective—is correct. Adopting the “new” view of dividendtaxation in FRB/US would likely have little effect on our qualitative conclusions. For example, the first simu-lation reported below induces a 10 percent drop in equity prices; the formula and parameters in Auerbach(1996) imply that this drop would be about 8 1/2 percent under the new view.

13 This statement and the FRB/US equation for uncovered interest parity ignore the effects of taxes. Thoseeffects are very complex and, under some reasonable assumptions, of second–order importance.


364

eign indebtedness. The model also in-cludes a foreign sector that accounts formovements in the output gap, interestrates, and inflation rates of our major trad-ing partners.

International Trade and Capital Flows

Exports of goods and services in FRB/US depend primarily on foreign GDP, theexchange rate, and relative price levels;imports are a function of domestic GDP,the exchange rate, and relative price lev-els. The elasticities of both exports andimports with respect to the exchange rateand relative prices are estimated to be 0.7.These elasticities apply to trade in goodsand services together; because services areless price sensitive, the correspondingimplicit elasticities for trade in goods(merchandise trade) are close to unity.This figure roughly matches the averageelasticity for merchandise imports andexports estimated by Hooper, Johnson,and Marquez (1998), as well as the elas-ticities those authors report from two ear-lier “representative studies” (Cline, 1989;and Marquez, 1990).

The sum of net exports and net invest-ment income from abroad (the balancein the current account) is, as a matter ofaccounting, exactly equal to net foreigninvestment. Note that net foreign invest-ment equals investment by domestic resi-dents in other countries less domestic in-vestment undertaken by foreign residents,so it is the negative of the net capital in-flow from abroad in the balance of pay-ments. Thus, any change in the net capi-tal inflow is mirrored by equal–sizedchanges in net exports and net investmentincome. If an increase in budget deficitsgives rise to larger capital inflows thathelp finance government borrowing, netexports must decline by the same amount.

An important uncertainty for thepresent analysis is the responsiveness ofinternational capital flows to changes infiscal policy. Because an increase in the

capital inflow goes hand–in– hand with adecrease in net exports, we can restate thisuncertainty about capital flows as uncer-tainty about two factors: the response ofthe real exchange rate, and the elasticityof exports and imports with respect to thereal exchange rate. If we maintain our as-sumption that the value of the dollar isdetermined by uncovered interest parityin the short run and purchasing powerparity in the long run, then a greater netcapital inflow can occur only with greaterprice elasticities for exports and imports.Therefore, one of our simulations assumesthat the price elasticities of exports andimports are twice as large as in the stan-dard FRB/US specification—values thatare above the upper end of the range ofestimates found by previous researchers.Alternatively, we could drop our assump-tion about uncovered interest parity andpurchasing power parity, allowing fiscalpolicy to permanently alter the value ofthe dollar. In this case, a greater net capi-tal inflow would be consistent with themodel’s baseline price elasticities for im-ports and exports, as long as the dollarappreciated substantially in the short runand did not revert to its original level evenafter several decades. However, the ongo-ing capital flows that would occur underthis scenario would leave foreign portfo-lios with a rising share of dollar–denomi-nated assets; such a combination of dol-lar appreciation and a sustained increasein foreign indebtedness seems unlikely onboth logical and empirical grounds, as weexplain in the next section, so we do notpursue it further.

FISCAL POLICY AND FINANCIALMARKETS

In the simulations we present below, anincrease in the government budget defi-cit owing to lower taxes or higher govern-ment spending leads to higher interestrates, all else equal. This connection willnot surprise most readers, but neither is


365

it universally accepted, so we think thatthe subject merits discussion.

We begin with some national incomeaccounting identities. Let Y denote na-tional income, C private consumption, Sprivate saving, and T taxes less govern-ment transfer payments. The privatesector’s budget constraint implies that: Y= C + S + T. National income also equalsnational output, which can be divided intofour types of spending: Y = C + I + G +NX, where I is domestic investment, G isgovernment purchases of goods and ser-vices, and NX is net exports of goods andservices. Balance in the international ac-counts requires: NX = NFI, where NFI isnet foreign investment. Combining theseidentities yields an expression for equilib-rium in the market for loanable funds:S + (T – G) = I + NFI, where the left sideshows national saving as the sum of pri-vate and public saving, and the right sideshows the uses of these saved funds forinvestment at home and abroad.

Now consider the effect of a reductionin taxes. Unless the resulting drop inpublic saving is entirely offset by higherprivate saving and lower net foreign in-vestment (that is, larger capital inflows),domestic investment must decline.Achieving the decline in domestic invest-ment requires a higher interest rate. Inother words, a budget deficit that is notentirely offset by higher private savingand larger net inflows of foreign capitalwill raise the interest rate that equilibratessaving and investment. The key issue,then, is the size of these offsets. The FRB/US specification is consistent with ourreading of the voluminous empirical evi-dence in these areas: Both offsets are sub-stantial, but they are far from complete.

Private saving is likely to increase inresponse to a larger budget deficit becauseforward–looking consumers realize thatadditional government debt will force

higher future taxes. In an extreme case,households will save the entire tax cut tomeet their upcoming tax liability, and na-tional saving will be unchanged (Barro,1974). Elmendorf and Mankiw (1999) re-view the theoretical and empirical debateover the Ricardian equivalence proposi-tion and conclude that “a substantial frac-tion of households probably do not be-have as the proposition assumes” (page1654). In particular, they argue that con-sumption is not smoothed over timenearly as much as would be required forRicardian equivalence (or even the basiclifecycle model) to hold. For evidence onthis point, see Campbell and Mankiw(1989), Carroll and Summers (1991),Parker (1999), Souleles (1999), Wilcox(1989), and the survey by Browning andLusardi (1996).

Net capital inflows are also likely to in-crease in response to a larger budget defi-cit, but by less than one might expect in aworld where many countries allow capi-tal to move freely across their borders.Although gross capital flows are large, theempirical evidence shows that capitalmobility on a net basis is limited, perhapsbecause of asymmetric information or riskdiversification.14 One form of evidence isthe high correlation between domesticsaving and domestic investment, docu-mented by Feldstein and Horioka (1980),Obstfeld and Rogoff (2000), and others.Confirming evidence comes from signifi-cant differences in real interest rates acrosscountries (Mishkin, 1984, and Cumby andMishkin, 1986) and from the heavy con-centration of individual portfolios in do-mestic securities (French and Poterba,1991).

Elmendorf and Liebman (2000) con-clude from this set of research that a one–dollar increase in the budget deficit mightgenerate a 25–cent increase in private sav-ing, and that the resulting 75–cent de-

14 Mussa and Goldstein (1993), Gordon and Bovenberg (1996), and Obstfeld and Rogoff (2000) review the evidenceregarding international capital mobility and discuss a number of explanations for the observed immobility.


366

crease in national saving might generatea 25–cent increase in the net inflow of capi-tal; altogether, this leaves a 50–cent de-crease in domestic investment. Recentanalyses by the Congressional BudgetOffice (CBO, 1997 and 1999) assume thatthe increase in private saving would be40 cents and that the drop in domestic in-vestment would be only about 35 cents.In FRB/US, a one–dollar increase in thebudget deficit increases private saving byroughly 50 cents after several years, andraises net capital inflows by about 35cents, leaving only a 15–cent drop in do-mestic investment. Thus, the amount bywhich budget deficits crowd out domes-tic investment in FRB/US is considerablysmaller than the amount assumed in someother analyses.

It bears emphasis that larger responsesof private saving and international capi-tal flows to higher budget deficits will trimthe effect of deficits on interest rates, butmay also diminish the stimulative effectof deficits on the economy as a whole.Large responses do imply that less domes-tic investment is crowded out than other-wise—but they also imply that consump-tion rises by less and net exports fall bymore. We explore these issues in the simu-lations below.

Notwithstanding the indirect evidencediscussed earlier in this section, directevidence for the proposition that largerbudget deficits lead to higher interestrates is admittedly weak. Indeed,Elmendorf and Mankiw summarize theresearch on reduced–form relationshipsbetween these variables by stating that “ithas typically supported the Ricardianview that budget deficits have no effecton interest rates” (page 1658). How doesone reconcile this (lack of) reduced–formevidence with the indirect evidence onhousehold behavior and capital mobilitythat implies a strong connection betweendeficits and interest rates? Elmendorf andMankiw argue that the reduced–form lit-erature:

“. . . is ultimately not very informative.Examined carefully, the results are sim-ply too hard to swallow, for three reasons.First, the estimated effects of policy vari-ables are often not robust to changes insample period or specification. Second,the measures of expectations included inthe regressions generally explain only asmall part of the total variation in inter-est rates . . . Third, [these papers often]cannot reject the hypothesis that govern-ment spending, budget deficits, and mon-etary policy have no effect on interest rates. . . These findings suggest that this frame-work has little power to measure the trueeffects of policy” (page 1658).

Uncovering the effect of fiscal policy oninterest rates is difficult because so manyeconomic variables affect those rates.Elmendorf and Mankiw emphasize:

“Interest rates depend on expectations offiscal policy and other variables, and thoseexpectations are hard to measure. A num-ber of studies use forecasts from vectorautoregressions as a proxy for expecta-tions, [which] assumes that variables fol-low a stable time–series process and donot incorporate non–quantitative infor-mation . . . Measurement error in the prox-ies for expectations biases the estimatedcoefficients toward zero and, thus, towardthe null hypothesis of Ricardian equiva-lence” (page 1657).

In sum, we think that the positive re-sponse of interest rates to increases in fu-ture budget deficits displayed in FRB/US(all else equal) is an unsurprising andpositive aspect of the model.

EFFECTS OF A SUSTAINED CUT INPERSONAL INCOME TAXES

In this section we consider the effectsof a sustained (but not completely perma-nent) cut in personal income taxes undera variety of different assumptions; thesealternatives highlight the channels


367

through which fiscal policy operates andillustrate the effect of variations in themodel. The baseline for all of the simula-tions is the economy in steady–state equi-librium, and all variables are described andshown in terms of changes relative to thatbaseline; for concreteness, we label the ini-tial period for the policies as the first quar-ter of 2002. Except when noted, we assumethat the policy we describe is perfectly cred-ible to households, businesses, and finan-cial–market participants. This assumptionis an important one, and allowing policiesto win credibility only gradually would bean interesting topic for future research.

Standard FRB/US Specification

We assume that tax rates are cut byenough to reduce federal personal incometaxes by 1 percent of GDP, or roughly $100billion in 2002, under static assumptionsabout the tax base. This reductionamounts to slightly more than a 10 per-cent cut in the average federal income taxrate. We further assume that the tax cutstems from a proportional reduction inmarginal tax rates across the income dis-tribution, so that all workers experienceroughly a 10 percent drop in their mar-ginal tax rates. Of course, this assumptionstrengthens the labor–supply effect rela-tive to more realistic alternatives in whichpart or all of the tax cut arises from infra-marginal changes.15 We leave the tax cutin place for ten years, at the end of whichthe debt–GDP ratio (incorporating theextra debt owing to higher interest costs)is higher by about 13 percentage points.

The tax rate then rises gradually—firstto hold the increase in the budget deficitrelative to baseline at 2 percent of GDPfor several years, and then to return thebudget to balance. This policy causes thedebt–GDP ratio to peak at 25 percentagepoints above the baseline after 20 years,an increment that roughly matches the in-crease in this ratio between 1980 and1993.16

The macroeconomic effects of thispolicy are summarized in Figure 1 and thetop part of Table 1. (Additional resultsfor this simulation and the subsequentones are provided in appendix tablesavailable from the authors upon request.)The top left panel of the figure shows thefederal budget surplus as a percent ofGDP between 2001 and 2006. The top rightpanel shows that the federal funds rateinches up over time. The same panelshows that the 10–year Treasury yieldjumps up when the tax cut is announced,anticipating future increases in the fundsrate—which occur largely after 2006 andtherefore do not appear in this figure.17

The middle left panel shows that equityprices drop nearly 10 percent when the taxcut is announced, owing to both a higherdiscount rate on future dividends (be-cause of higher interest rates) and lowerexpected dividends (because of the de-cline in business investment). The middleright panel shows that real consumptionincreases, as the effect of the tax cut onpermanent income more than offsets theeffect of lower wealth, and that real invest-ment is crowded out by higher real inter-est rates.

15 CBO (2001) reports that the average marginal tax rate on wages and salaries is 24 percent. Combined with asocial insurance tax rate of roughly 15 percent (combined employer and employee share) and a state and localincome tax rate of 5 percent, the increase in the after–tax wage stemming from the assumed cut in the federalmarginal rate equals: ((1 – .05 – .15 – .213)/(1 – .05 – .15 – .24)) – 1 = 4.8 percent. With our assumed elasticity oflabor supply equal to 0.17, this tax cut would induce an increase in potential labor supply of roughly 0.8percent and an increase in potential GDP of about 0.5 percent.

16 Allowing the tax cut to persist for several additional decades would not appreciably alter the short–run re-sponse of the economy to the policy change.

17 Those future funds–rate increases owe to two factors: a rise in the equilibrium real rate over the followingfifteen years that is implied by the relative effect of fiscal policy on aggregate demand and aggregate supply;and a somewhat higher average rate of inflation going forward.


368

The bottom left panel of the figureshows that the level of potential realGDP increases about one–half percentover two years, with the additional laborsupply partly offset by diminished capi-tal services due to weaker investment. Ac-tual real GDP rises not quite one–half per-cent after a year, and is aligned quite

closely with potential GDP after a coupleof years.

The bottom right panel shows that theunemployment rate dips about one–tenthpercentage point within a few quarters,because the increased labor demand fromhigher current and expected future out-put more than offsets the increased labor

Figure 1. Effect of a Sustained Cut in Personal Income Taxes Equal to 1% of GDP


369

supply from lower marginal tax rates.18

The inflation rate drifts up slightly owingto expectations of lower unemploymentin the future.

The top part of Table 1 presents thechanges in saving and investment underthis scenario. Looking at the valuesfor 2006, one can see that private savingincreases by about 50 percent of thedrop in public saving (.71/–1.38), andthat international capital flows offsetabout 70 percent of the drop in nationalsaving (–.49/[.71–1.38]). Altogether,domestic investment falls by about 15percent of the increase in the budgetdeficit.

Figure 2 summarizes the long–run ef-fect of this tax cut. Over time, the reduc-tion in capital services increases in mag-nitude, and the labor supply effect turnsnegative because higher taxes are neededto service the additional debt while bal-ancing the budget. As a result, potentialGDP falls below baseline after about 15years, and actual GDP falls with it.

In sum, simulating the FRB/US modelsuggests that a sustained tax cut reducesoutput in the long run and raises outputby less than 50 cents per dollar of tax re-duction in the short run. The fact that themultiplier is relatively small is arguablynot an important concern: As long as the

TABLE 1EFFECTS OF FISCAL POLICY ON GROSS SAVING AND INVESTMENT(Q4 CHANGE FROM BASELINE, EXPRESSED AS A PERCENT OF GDP)

2002 2003 2004 2005 2006

Sustained income–tax cut worth 1 percent of GDP on static basisPrivate saving .67 .63 .64 .67 .71Government saving –1.01 –1.10 –1.19 –1.29 –1.38Domestic investment –.13 –.19 –.19 –.19 –.22Net foreign investment –.22 –.28 –.36 –.43 –.49

Income–tax cut with more rule–of–thumb consumptionPrivate saving .43 .53 .60 .65 .70Government saving –.87 –1.08 –1.19 –1.29 –1.40Domestic investment –.11 –.25 –.24 –.24 –.26Net foreign investment –.33 –.30 –.35 –.40 –.44

Income–tax cut with greater international capital mobilityPrivate saving .57 .42 .41 .46 .50Government saving –.88 –.99 –1.09 –1.19 –1.30Domestic investment –.08 –.13 –.12 –.09 –.07Net foreign investment –.23 –.43 –.56 –.64 –.72

10 percent investment tax credit for equipment and softwarePrivate saving .33 .66 .92 1.14 1.33Government saving –1.07 –1.22 –1.42 –1.69 –2.00Domestic investment .15 .70 1.02 1.13 1.12Net foreign investment –.89 –1.25 –1.51 –1.68 –1.79

Sustained increase in government purchases worth 1 percent of GDP on static basisPrivate saving .48 .58 .65 .68 .68Government saving –.94 –1.11 –1.25 –1.37 –1.50Domestic investment –.13 –.24 –.28 –.33 –.38Net foreign investment –.32 –.29 –.32 –.37 –.43

18 In this and other simulations that include a labor supply response to fiscal policy, the unemployment rate ispushed up initially by the lags in firms’ decisions to hire newly available workers. To the extent that greaterlabor supply makes the unemployment rate higher than it would otherwise be, that increment should not beinterpreted as a harmful effect of the policy. On the contrary, reducing the disincentive to work presumablyincreases economic efficiency and (all else equal) well–being, even if the additional labor supply becomesemployed only gradually.


370

multiplier has the customary sign, anydesired amount of economic stimuluscould still be generated by a sufficientlylarge tax cut. However, tax policy aims toachieve multiple objectives in addition tomacroeconomic stabilization—such as ef-ficiency and equity, both within and acrosscohorts. These goals may be adversely

affected by substantial changes in taxes,especially if such changes are promptedby cyclical developments that are difficultto predict in advance. Therefore, if mac-roeconomic stabilization requires large,unpredictable swings in taxes, fiscalpolicy is relatively less attractive as a sta-bilizing tool.

Figure 2. Longer–Run Effect of a Persistent Cut in Personal Income Taxes Equal to 1% of GDP


371

Gradual Financial–Market Response

A principal motivation for this paper isthe argument that forward–looking finan-cial markets undo much of the putativestimulative effect of lower taxes or higherspending. To illustrate the effect of for-ward–looking financial markets, our sec-ond simulation assumes that long–terminterest rates do not jump up when thetax cut is introduced, but converge gradu-ally (at a rate of 10 percent per quarter)toward the values consistent with futureshort–term interest rates. Equity pricesand the value of the dollar converge totheir fully rational values at a similarlyslow speed. All other expectational vari-ables are the same as in our first simula-tion, so this scenario embodies the im-probable combination of myopic investorsand fully rational households and firms;however, we emphasize that this simula-tion is not intended as a realistic possibil-ity, but as an artificial experiment to quan-tify the importance of forward–looking fi-nancial markets to our previous result.

Figure 3 presents the results of thissimulation. GDP rises by nearly 1 percentafter a year, compared with less than one–half percent when financial markets areforward–looking. By this metric, the for-ward–looking nature of financial marketsoffsets roughly 50 percent of the incipientstimulative effect of this tax cut. Becausethe boom in output is substantially largerthan the slight gain in potential output,the unemployment rate falls nearly half apercentage point, and inflation picks up abit. The federal funds rate rises about 1percentage point in response to the in-crease in inflation and the increase in ac-tual output in excess of potential GDP.

An alternative type of “gradual re-sponse” is one by which monetary policyreacts to economic developments with asubstantial lag. However, when we repeatour first simulation holding the federalfunds rate fixed for two years, and set ac-cording to the Taylor rule thereafter, we

find essentially no difference in the effectof the tax cut on real GDP. The key pointis that financial–market participants areassumed to understand that the funds ratewill eventually be raised, so long–terminterest rates increase immediately and byroughly as much as in the first simulation.Thus, the speed of the monetary policyreaction is less important than the speedof realization by financial markets that areaction will ultimately ensue.

Circumstances may arise under whichthis anticipatory bond–market responsewould not operate in the usual manner.In particular, if short–term interest ratesare expected to remain at zero for a pro-longed period—perhaps the current situ-ation in Japan—then stimulative fiscalactions may not produce the standard in-crease in bond yields. Thus, extreme con-ditions exist under which monetary policybecomes relatively ineffective as a stabili-zation tool, and fiscal policy proves moreuseful. For further discussion of this is-sue, see Reifschneider and Williams (2000)and Clouse, Henderson, Orphanides,Small, and Tinsley (2000).

Holding Potential Output Fixed

Analyses of countercyclical fiscal policygenerally focus on the effects of policy onaggregate demand and do not considertheir effects on aggregate supply. A no-table exception is Auerbach and Feenberg(2000), who argue that the stabilizing ef-fect of the progressive income tax operat-ing through a wage–responsive supplyof labor may be as important quantita-tively as its traditional stabilizing effectthrough aggregate demand. In particular,Auerbach and Feenberg estimate that acyclical decline in income reduces mar-ginal tax rates sufficiently to induce amarked temporary increase in labor sup-ply. To illustrate the impact on our resultsof changes in aggregate supply, our fourthsimulation holds potential output fixed.


372

Figure 4 shows that real GDP risesabout one–tenth of a percent in the firstquarter and then returns to baselinewithin several years, compared with amore sustained pickup closer to one–halfpercent in our first simulation. Thus,the response of aggregate supply—laborsupply, in particular—accounts for morethan half of the total stimulative effect of

the tax cut. Recall, however, that our as-sumed tax cut stems entirely from a re-duction in marginal tax rates; if we hadassumed some reduction in the inframar-ginal tax burden instead, the total stimu-lative effect of the tax cut would besmaller, and the share stemming fromchanges in labor supply would be smalleras well.

Figure 3. Effect of a Sustained Cut in Personal Income Taxes with a Gradual Financial Market Response


373

More Rule–of–Thumb Consumption

The next simulation substantially in-creases the sensitivity of consumer spend-ing to current income. Under the baselineFRB/US specification, a one–dollar per-manent shock to income raises spendingby about 20 cents initially; under this al-ternative specification, the same shock

boosts spending by about 50 cents, in linewith Campbell and Mankiw’s (1989) esti-mate of the share of consumption done byrule–of–thumb consumers.

Figure 5 shows that the tax cut providesa larger boost to economic activity underthese conditions. Compared with thefirst simulation, consumption naturallyrises more abruptly, and investment

Figure 4. Effect of a Sustained Cut in Personal Income Taxes Holding Potential Output Fixed


374

falls by about the same amount. Thus,real GDP now increases over three–quar-ters of a percent after several quarters,compared with less than one–half percentin the first simulation, and the unemploy-ment rate falls about three–tenths of apercentage point, compared with one–tenth in the first simulation. The second

part of Table 1 shows that private savingrises much less during the first yearthan under the standard model, butincreases roughly in line with the standardmodel in subsequent years. All told,the short–run impact of the tax cut is stillfairly small, with a multiplier less thanone.

Figure 5. Effect of a Sustained Cut in Personal Income Taxes with More Rule–of–Thumb Consumption


375

Greater Reaction of InternationalCapital Flows

The next simulation increases the netinflow of foreign capital in response to thetax cut. As described earlier, we model thisgreater reaction by doubling the price elas-ticity of exports and imports. The thirdpart of Table 1 shows that internationalcapital inflows offset nearly 100 percentof the drop in national saving (–.72/[.50 –1.30]) in this scenario.

Figure 6 shows that the ten–year Trea-sury yield initially rises less than 10 basispoints in this simulation, compared withabout 25 basis points in the first simula-tion; most of this small increase reflects anuptick in expected average inflation overthe next ten years. Because real interestrates are barely changed, so is investment.However, the larger capital inflow requiresa larger decline in net exports; relative tothe first simulation, more of the increasein consumption demand is satisfied byforeign production, and less by a shift indomestic production from investmentgoods to consumption goods. Net exportsrespond gradually to movements in theexchange rate, and therefore decline lessthan consumption rises; on balance, realGDP rises a little more than half a percentafter a year, a slightly larger effect than inthe first simulation. Yet, the short–run ef-fect of the tax cut is again small, with amultiplier of roughly one–half.

EFFECTS OF ALTERNATIVEFISCAL POLICIES

In this section, we consider the effectsof several alternative fiscal policies: aphased–in personal income tax cut, a one–time personal tax rebate, a tax incentivefor business investment, and an increasein government spending.

A Phased–In Personal Income Tax Cut

We turn first to the sustained cut in per-sonal income taxes examined in the previ-ous simulations, but assume now that itoccurs in ten equal annual increments start-ing in 2002 rather than occurring all at once.We also assume that the full policy is per-fectly credible from the beginning to house-holds, businesses, and financial–marketparticipants; all of the other assumptionsare identical to the first simulation.

Figure 7 shows that this phased–in taxcut slightly reduces real GDP for severalyears. The direct stimulus of lower taxesis delayed: Households discount expectedfuture income at a high rate, so tax reduc-tions more than a few years away haveonly a modest effect on current consump-tion, all else equal.19 Financial markets donot discount the future as heavily, so thedamping influences of higher interestrates and lower equity prices affect theeconomy immediately. The opposing in-fluences of higher after–tax income andlower wealth leave consumption littlechanged for the first three or four years,while investment falls noticeably from thestart. Not until 2006, when half of the fulltax cut is in place, does aggregate demandrise above its baseline value.

One–Time Personal Tax Rebate

We now turn to the effects of a one–time$100 billion rebate on personal incometaxes—an amount roughly equal to the firstyear of the sustained tax reduction, but paidout in one quarter. This rebate is assumedto involve no change in marginal tax ratesand thus induce no change in labor supply.

Figure 8 shows that this policy boostsconsumption by about one–half percentin the quarter when the rebate is issued,compared with a rebate equal to 4 percent

19 We assume that households do not substitute their desired labor supply intertemporally. Therefore, theirlabor supply in the early years of the tax cut is boosted by the reduction in marginal tax rates that has alreadyoccurred, but it is not affected by the further declines in marginal rates ahead.


376

of GDP on an annualized basis in thatquarter. The rebate has virtually no effecton consumption in subsequent quarters.For the vast majority of households inFRB/US whose consumption respondsgradually to changes in permanent in-come, spending barely reacts to the rebateat all; for the small minority of households

who set consumption equal to current in-come, spending jumps in the quarter ofthe rebate and recedes in the next quarter.Long–term interest rates and investmenthardly budge. As a result, GDP is about0.3 percent higher in the quarter when therebate is issued and virtually unchangedin subsequent quarters.

Figure 6. Effect of a Sustained Cut in Personal Income Taxes with Greater Reaction of InternationalCapital Flows


377

Figure 9 presents the effects of the samepolicy under the alternative specificationthat about half of consumption is done byrule–of–thumb consumers, who are as-sumed to consume their entire rebatewithin two quarters. As one would expect,the tax rebate has a much larger effect onconsumption and output under these cir-cumstances. Investment rises a bit

through an accelerator effect and thendrops slightly below baseline. GDP ismore than 1 percent higher for the firsthalf year, then below baseline for abouttwo years, and back at baseline thereaf-ter. The unemployment rate follows theopposite pattern: it is down nearly half apercentage point initially, above baselinefor a while, and then back at baseline. This

Figure 7. Effect of a Sustained Cut in Personal Income Taxes Phased In Over 10 Years


378

oscillation of output and unemploymentstems from the lagged effects of the in-crease in the federal funds rate inducedby the initial surge in output.

In sum, a tax rebate appears to impart asubstantial short–run stimulus to theeconomy if a significant share of house-holds spend the money. How households

actually respond to temporary tax cuts isunclear. Shapiro and Slemrod (2001) re-port that only a fifth of households in theMichigan Survey Research Center’s Sur-vey of Consumers said that they expectedto spend last year’s tax rebate. However,one wonders whether the drumbeat ofnational exhortation to increase saving

Figure 8. Effect of a One–Time $100 Billion Personal Income Tax Rebate


379

had persuaded households that saving therebate was the “right” thing to do, lead-ing them to respond to the question morein line with their aspirations than theirbehavior. In any event, a number of stud-ies find that consumption is substantiallymore sensitive to current income than thebasic lifecycle model predicts, as noted

above. Some studies suggest in particu-lar that households have a fairly high pro-pensity to consume out of transitory taxchanges: Shapiro and Slemrod (1995) findthat about 40 percent of householdsplanned to spend the extra cash flow fromthe 1992 change in tax withholding (butnot tax liability), and Souleles (1999) esti-

Figure 9. Effect of a One–Time $100 Billion Personal Income Tax Rebate with More Rule–of–ThumbConsumption


380

mates that between one–third and two–thirds of every dollar of tax refund isspent within one quarter. On the otherhand, analyses of aggregate consumptionfollowing the 1975 tax rebate (Blinder,1981; Modigliani and Steindel, 1977;and Poterba, 1988) find somewhat smallerinitial effects of the rebate on consump-tion.

Tax Incentive for Business Investment

An alternative tax change often consid-ered as a short–run economic stimulus isa tax incentive for business investment.Our next simulation shows the macroeco-nomic effects of granting a 10 percent in-vestment tax credit (ITC) on equipmentand software. We assume that the ITC re-

Figure 10. Effect of a 10 Percent Investment Tax Credit on Equipment and Software


381

mains in place for a sustained period andthat all economic agents believe that to betrue. Clearly, these are limiting assump-tions.20 The magnitude of our simulatedITC is consistent with U.S. tax policy be-

tween the mid–1970s and mid–1980s; itreduces the user cost of equipment capi-tal by roughly 14 percent.

Figure 10 shows that a credible, sus-tained investment tax credit provides a

Figure 11. Effect of a Sustained Increase in Federal Expenditures Equal to 1 Percent of GDP

20 Our original intention in this study was to include an analysis of a temporary ITC as well, but incorporatingthe effects of temporary tax policies on the user cost of capital and business investment turned out to be moredifficult than we had anticipated. The root of the problem is that the effect of temporary tax incentives has not


382

substantial stimulus to economic activity—but with a lag of several quarters. The ex-planation for this lag is simply that adjust-ment costs make investment respond onlyslowly to the jump in the desired long–runcapital stock caused by the ITC. Meanwhile,financial markets foresee the coming in-crease in real GDP and acceleration of in-flation, and they anticipate a significantincrease in the federal funds rate; as a re-sult, long–term interest rates rise immedi-ately by more than one percentage point.Equity prices drop 15 percent, because theeffect of higher future dividends (stemmingfrom lower corporate taxes and higher in-vestment) is more than offset by the effectof the higher interest rates used to discountthose dividends. The fourth part of Table 1shows that the increase in investment is fi-nanced by a much larger inflow of foreigncapital than in our other simulations.

Increase in Government Spending

Our final simulation examines the ef-fects of a sustained increase in federalpurchases equal to 1 percent of GDP. Fig-ure 11 shows that this policy raises GDPby nearly one percentage point during thefirst several quarters, compared with aboost to GDP of less than one–half per-centage point from the same–sized reduc-tion in personal income taxes shown inFigure 1. Changes in government spend-ing have a larger stimulative effect thanchanges in taxes because a sizable shareof each dollar of lower taxes goes toprivate saving, whereas each dollar ofadditional government spending boostsaggregate spending by the full dollar.Consumption falls below baseline in thisscenario, and investment drops moresharply than in the first simulation.

These results would appear to suggestthat an increase in government outlays isan effective way to stimulate the economyin the short run. However, our calculationsignore a key practical question, which is thespeed at which outlays can be increased.Passing appropriations bills takes time, anddisbursing additional appropriated fundscan take even longer, depending on the typeof spending being considered. Thus, a com-parison of Figures 1 and 10 may overstatethe advantage of higher spending relativeto lower taxes as a short–run stimulus.

CONCLUSION

Using an empirical, open–economy, ra-tional–expectations macromodel, we ex-amined the short–run effects of a varietyof fiscal policies. We found that tax cuts andspending increases generally stimulateeconomic activity in the short run, contraryto the extreme view that forward–lookingfinancial markets more than offset the di-rect expansionary impulse of those actions.However, the magnitude of the stimulusis greatly attenuated by the financial–mar-ket feedback, and many policies have littlenet effect on output in the short run. Forexample, a sustained cut in personal in-come taxes raises real GDP by less than theamount of the tax cut itself, and it likelyreduces GDP if phased in gradually overtime. A tax rebate boosts the economy to anoticeable extent only if a significant shareof consumption tracks current income, anda sustained tax incentive for business in-vestment boosts the economy to a signifi-cant degree only after several quarters.21 Asustained increase in government pur-chases raise output quickly, with a multi-plier close to one, but may be difficult toimplement within a few quarters.

been fully resolved, although work on this problem is ongoing (Cohen, Hansen, and Hassett, 2002). We would alsohave liked to analyze the effects of imperfect credibility, because the history of U.S. tax policy shows that perma-nent investment incentives do not exist. However, this analysis too would have required a persuasive, empiri-cally–based model of the effect of temporary investment incentives, which FRB/US does not presently include.

21 See Congressional Budget Office (2002) for a discussion of the relative stimulus provided by various other taxpolicies.


383

Acknowledgments

This paper was prepared for theNational Tax Association’s 2002 SpringSymposium. We thank Darrel Cohen, Ja-son Cummins, Rochelle Edge, GlennFollette, Bill Gale, Kevin Hassett, DaveStockton, Sandy Struckmeyer, StaceyTevlin, and David Wilcox for helpful com-ments. The views expressed are our ownand not necessarily those of the FederalReserve Board or other members of itsstaff.

REFERENCES

Alesina, Alberto, and Roberto Perotti.“Fiscal Adjustments in OECD Countries:Composition and Macroeconomic Effects.”International Monetary Fund 44 No. 2 (June,1997): 210–48.

Auerbach, Alan J.“Share Valuation and Corporate EquityPolicy.” Journal of Public Economics 11 No. 3(June, 1979): 291–305.

Auerbach, Alan J.“Tax Reform, Capital Allocation, Efficiency,and Growth.” In Economic Effects of Funda-mental Tax Reform, edited by Henry J. Aaronand William G. Gale, 29–81. Washington,D.C.: Brookings Institution, 1996.

Auerbach, Alan J., and Daniel Feenberg.“The Significance of Federal Taxes as Auto-matic Stabilizers.” Journal of Economic Per-spectives 14 No. 3 (Summer, 2000): 37–56.

Barro, Robert J.“Are Government Bonds Net Wealth?” Jour-nal of Political Economy 82 No. 6 (November,1974): 1095–117.

Blanchard, Olivier J.“Current and Anticipated Deficits, InterestRates and Economic Activity.” European Eco-nomic Review 25 No. 1 (June, 1984): 7–27.

Blanchard, Olivier J., and Roberto Perotti.“An Empirical Characterization of the Dy-namic Effects of Changes on GovernmentSpending and Taxes on Output.” Massachu-setts Institute of Technology. Mimeo, No-vember 2001.

Blinder, Alan D.“Temporary Income Taxes and ConsumerSpending.” Journal of Political Economy 89No. 1 (February, 1981): 26–53.

Bradford, David.“The Incidence and Allocation Effects of a Taxon Corporate Distributions.” Journal of Pub-lic Economics 15 No. 1 (April, 1981): 1–22.

Branson, William.“Causes of Appreciation and Volatility of theDollar.” The U.S. Dollar–Recent Developments,Outlook, and Policy Options. Federal ReserveBank of Kansas City, 1985.

Brayton, Flint, and Peter Tinsley.“A Guide to FRB/US: A MacroeconomicModel of the United States.” Federal Re-serve Board Finance and Economics Discus-sion Series, 1996.

Brayton, Flint, Andrew Levin, Ralph Tryon,and John C. Williams.

“The Evolution of Macro Models at the Fed-eral Reserve Board.” Carnegie–Rochester Con-ference Series on Public Policy 47 No. 0 (De-cember, 1997): 43–81.

Brayton, Flint, Eileen Mauskopf, DavidReifschneider, Peter Tinsley, and John Williams.

“The Role of Expectations in the FRB/USMacroeconomic Model.” Federal Reserve Bul-letin (April, 1997): 227–45.

Browning, Martin, and Annamaria Lusardi.“Household Saving: Micro Theories andMicro Facts.” Journal of Economic Literature,34 No. 5 (December, 1996): 1797–855.

Campbell, John Y., and N. Gregory Mankiw.“Consumption, Income and Interest Rates:Reinterpreting the Time Series Evidence.”NBER Macroeconomics Annual (1989): 185–216.

Carroll, Christopher D.“A Theory of the Consumption FunctionWith and Without Liquidity Constraints.”Journal of Economic Perspectives 15 No. 3(Summer, 2001): 23–45.

Carroll, Christopher D., and Lawrence H.Summers.

“Consumption Growth Parallels IncomeGrowth: Some New Evidence.” In National Sav-ing and Economic Performance, edited by B. Dou-glas Bernheim and John B. Shoven, 305–47.Chicago: University of Chicago Press, 1991.


384

Cline, William.United States External Adjustment and theWorld Economy. Washington, D.C.: Institutefor International Economics, 1989.

Clouse, James, Dale Henderson, AthanasiosOrphanides, David Small, and Peter Tinsley.

“Monetary Policy When the Nominal Short–Term Interest Rate is Zero.” Federal ReserveBoard Finance and Economics DiscussionSeries, 2000.

Cohen, Darrel, and Glenn Follette.“The Automatic Fiscal Stabilizers: QuietlyDoing Their Thing.” Economic Policy Review(April, 2000): 35–67.

Cohen, Darrel, Dorthe–Pernille Hansen, andKevin Hassett.

“The Effects of Temporary Partial Expens-ing on Investment Incentives in the UnitedStates.” National Tax Journal 55 No. 3 (Sep-tember, 2002): 457–66.

Council of Economic Advisers.Economic Report of the President. Washington,D.C.: Government Printing Office, 1984.

Council of Economic Advisers.Economic Report of the President. Washington,D.C.: Government Printing Office, 1994.

Cumby, Robert E., and Frederic S. Mishkin.“The International Linkage of Real InterestRates: The European–U.S. Connection.”Journal of International Money and Finance 5No. 1 (March, 1986): 5–23.

Edelberg, Wendy, Martin Eichenbaum, andJonas D. M. Fisher.

“Understanding the Effects of a Shock toGovernment Purchases.” NorthwesternUniversity. Mimeo, 1998.

Elmendorf, Douglas W., and Jeffrey B. Liebman.“Social Security Reform and National Sav-ing in an Era of Budget Surpluses.”Brookings Papers on Economic Activity 2(2000): 1– 71.

Elmendorf, Douglas W., Jeffrey B. Liebman,and David W. Wilcox.

“Fiscal Policy and Social Security Policy Dur-ing the 1990s.” In American Economic Policyin the 1990s, edited by Jeffrey A. Frankel andPeter R. Orszag, 61–119. Cambridge, MA:MIT Press, 2002.

Elmendorf, Douglas W., and N. GregoryMankiw.

“Government Debt.” In Handbook of Macro-economics, edited by John B. Taylor andMichael Woodford, 1615–69. North–Hol-land, Amsterdam: Elsevier Science, 1999.

Fatas, Antonio, and Ilian Mihov.“The Effects of Fiscal Policy on Consump-tion and Employment: Theory and Evi-dence.” INSEAD. Mimeo, 1999.

Feldstein, Martin, and Charles Horioka.“Domestic Savings and International Capi-tal Flows.” Economic Journal 90 No. 358 (June,1980): 314–29.

French, Kenneth R., and James M. Poterba.“Investor Diversification and InternationalEquity Markets.” American Economic Review81 No. 2 (May, 1991): 222–6.

Friedman, Milton A.A Theory of the Consumption Function.Princeton: Princeton University Press, 1957.

Friedman, Milton A.“Windfalls, the ‘Horizon,’ and Related Con-cepts in the Permanent Income Hypothesis.”In Measurement in Economics, edited by CarlChrist et. al., 1–28. Stanford: Stanford Uni-versity Press, 1963.

Gale, William G., and Samara R. Potter.“An Economic Evaluation of the EconomicGrowth and Tax Relief Reconciliation Act of2001.” National Tax Journal 55 No. 1 (March,2002): 133–86.

Giavazzi, Francesco, Tullio Jappelli, and MarcoPagano.

“Searching for Non–Linear Effects of FiscalPolicy: Evidence from Industrial and Devel-oping Countries.” NBER Working Paper No.7460. Cambridge, MA: National Bureau ofEconomic Research, 2000.

Gordon, Roger H., and A. Lans Bovenberg.“Why is Capital So Immobile Internationally?Possible Explanations and Implications forCapital Income Taxation.” American EconomicReview 86 No. 5 (December, 1996): 1057–75.

Hemming, Richard, Michael Kell, and SelmaMahfouz.

“The Effectiveness of Fiscal Policy in Stimu-lating Economic Activity—A Review of the


385

Literature.” International Monetary Fund,March, 2000.

Hooper, Peter, Karen Johnson, and JaimeMarquez.

“Trade Elasticities for G–7 Countries.” Fed-eral Reserve Board International FinanceDiscussion Papers, Number 609, 1998.

King, Mervyn.Public Policy and the Corporation. London:Chapman and Hall, 1977.

Marquez, Jaime.“Bilateral Trade Elasticities.” Review of Eco-nomics and Statistics 72 No. 1 (February,1990): 70–7.

Mishkin, Frederic S.“Are Real Interest Rates Equal Across Coun-tries? An Empirical Investigation of Inter-national Parity Conditions.” Journal of Fi-nance 39 No. 5 (December, 1984): 1345–57.

Modigliani, Franco, and Charles Steindel.“Is a Tax Rebate an Effective Tool for Stabi-lization Policy?” Brookings Papers on Eco-nomic Activity 1 (1977): 175–209.

Mountford, Andrew, and Harald Uhlig.“What are the Effects of Fiscal Policy Shocks?”University of London. Mimeo, 2002.

Mussa, Michael, and Morris Goldstein.“The Integration of World Capital Markets.”In Changing Capital Markets: Implications forMonetary Policy, 245–313. Kansas City: Fed-eral Reserve Bank of Kansas City, 1993.

Obstfeld, Maurice, and Kenneth Rogoff.“The Six Major Puzzles in International Mac-roeconomics: Is There a Common Cause?”NBER Macroeconomics Annual, 2000.

Parker, Jonathan.“The Reaction of Household Consumptionto Predictable Changes in Payroll TaxRates.” American Economic Review 89 No. 4(September, 1999): 959–73.

Poterba, James M.“Are Consumers Forward Looking? Evidencefrom Fiscal Experiments.” American EconomicReview 78 No. 2 (May, 1988): 959–73.

Poterba, James M., and Lawrence H. Summers.“The Economic Effects of Dividend Taxa-tion.” In Recent Advances in Corporate Finance,edited by E. Altman and M. Subrahmanyam,

227–84. Homewood, IL: Richard D. Irwin,1985.

Ramey, Valerie A., and Matthew D. Shapiro.“Costly Capital Reallocation and the Effectsof Government Spending.” Carnegie–Roch-ester Conference Series on Public Policy 48No. 0 (June, 1998): 145–94.

Reifschneider, David, Robert Tetlow, and JohnWilliams.

“Aggregate Disturbances, Monetary Policy,and the Macroeconomy: The FRB/USPerspective.” Federal Reserve Bulletin (Janu-ary, 1999): 1–19.

Reifschneider, David, and John Williams.“Three Lessons for Monetary Policy in aLow Inflation Era.” Journal of Money, Creditand Banking 32 No. 4, Part 2 (November,2000): 936–66.

Shapiro, Matthew D., and Joel Slemrod.“Consumer Response to the Timing ofIncome: Evidence from a Change in TaxWithholding.” American Economic Review 85No. 1 (March, 1995): 274–83.

Shapiro, Matthew D., and Joel Slemrod.“Consumer Response to Tax Rebates.” Uni-versity of Michigan. Mimeo, 2001.

Souleles, Nicholas.“The Response of Household Consumptionto Income Tax Refunds.” American EconomicReview 89 No. 4 (September, 1999): 947–58.

Summers, Lawrence H.“Treasury Secretary Remarks to the ConcordCoalition.” November 15, 2000.

Taylor, John B.“Discretion versus Policy Rules in Practice.”Carnegie–Rochester Conference Series on Pub-lic Policy 39 No. 0 (December, 1993): 195–214.

Taylor, John B.“Reassessing Discretionary Fiscal Policy.”Journal of Economic Perspectives 14 No. 3(Summer, 2000): 21–36.

U.S. Congressional Budget Office.An Economic Model for Long–Run BudgetSimulations. Washington, July, 1997.

U.S. Congressional Budget Office.The Long–Term Budget Outlook: An Update.Washington, D.C., December, 1999.


386

U.S. Congressional Budget Office.Description of CBO’s Models and Methods forProjecting Federal Revenues. Washington,D.C., 2001.

U.S. Congressional Budget Office.Economic Stimulus: Evaluating ProposedChanges in Tax Policy. Washington, D.C.,January, 2002 .

Wilcox, David W.“Social Security Benefits, Consumption Ex-penditure, and the Life Cycle Hypothesis.”Journal of Political Economy 97 No. 2 (April,1989): 288–304.

Short–Run Effects of Fiscal Policy with Forward–Looking ...€¦ · Short–Run Effects of Fiscal Policy with Forward–Looking Financial Markets 357 National Tax Journal Vol.

Documents