Money-Financed Fiscal Programs: A Cautionary Tale

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. Money-Financed Fiscal Programs: A Cautionary Tale William B. English, Christopher J. Erceg, and David Lopez-Salido 2017-060 Please cite this paper as: English, William B., Christopher J. Erceg, and David Lopez-Salido (2017). “Money- Financed Fiscal Programs: A Cautionary Tale,” Finance and Economics Discussion Series 2017-060. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2017.060. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

∗ Money-Financed Fiscal Programs: A Cautionary Tale William B. English Christopher J. Erceg David Lopez-Salido Yale University Federal Reserve Board Federal Reserve Board May 25, 2017 Abstract Anumberofprominenteconomistsandpolicymakershavearguedthatmoney-financedfiscal programs (helicopter drops) could be efficacious in boosting output and inflation in economies facing persistent economic weakness, very low inflation, and significant fiscal strains. We employ a fairly conventional macroeconomic model to explore the possible effects of such policies. While we do find that money-financed fiscal programs, if communicated successfully and seen as credible by the public, could provide significant stimulus, we underscore the risks that would be associated with such a program. These risks include persistently high inflation if the central bank fully adhered to the program; or alternatively, that such a program would be ineffective in providing stimulus if the public doubted the central bank’s commitment to such an extreme strategy. We also highlight how more limited forms of monetary and fiscal cooperation – such as a promise by the central bank to be more accommodative than usual in response to fiscal stimulus – may be more credible and easier to communicate, and ultimately more effective in providing economic stimulus. JEL Classification: E52, E58 Keywords: Monetary Policy, Fiscal Policy, Liquidity Trap, DSGE Model, Currency Union. ∗We thank Jim Clouse, Brian Doyle, Thomas Laubach, Brian Madigan, Dave Reifschneider, and Beth Anne Wilson for useful comments on an earlier draft. The opinions expressed here are solely those of the authors and do not necessarily reflect the views of the Board of Governors of the Federal Reserve System, the Federal Open Market Committee, or any members of their staffs.

1. Introduction A number of economists and policymakers (e.g., Buiter (2014) and Turner (2015)) have suggested that in a sufficiently dire economic situation, money-financed fiscal programs – sometimes referred to as “helicopter money” – could be used to provide stimulus beyond what can be accomplished with standard monetary policy alone. Such programs involve an increase in government spending or a reduction in taxation, with the fiscal spending financed directly by the central bank.1 These economists argue that such policies can be effective in stimulating economies suffering inflation that is persistently very low (or negative) and output that is far below potential, even if monetary policy is constrained by the effective lower bound on interest rates and fiscal policy is constrained by high levels of government debt. Our paper uses a New Keynesian model to examine the possible costs and benefits of coordinated monetary and fiscal policies. While we devote particular attention to money-financed fiscal expansions, we also consider more limited forms of coordination that are consistent with relatively transient departures from the usual central bank reaction function. We discuss the possible effects of such programs and some of the practical considerations that would likely complicate their use. We begin by considering a “textbook”’ money-financed fiscal program that involves an increase in government spending funded by an equal-sized permanent expansion in the stock of currency.2 A critical feature of such a program is that the central bank commits not to reverse the increase in currency in the future through monetary policy actions, so that the fiscal expansion is financed entirely by the inflation tax. In our benchmark model in which agents fully understand this commitment and regard it as credible, a money-financed program is a powerful tool for boosting output 1 The term “money-financed fiscal program” is from Bernanke (2016). Sometimes these programs are referred to as “helicopter drops” of money, though we think it is useful to bear in mind that the policies involve a mix of traditional fiscal and monetary policy actions. For recent discussion of such programs, see Turner (2015), Buiter (2014),andGal´ı(2016). Foracriticalperspective,seeBorioetal(2016)andCecchettiandSchoenholtz(2016). See Friedman (1969) for the original discussion of “helicopter money.” 2 Since we assume that the central bank can pay interest on reserves, currency is the only non-interest-bearing liability of the central bank. 1

and inflation. The government spending multiplier is around five even outside of a liquidity trap, so that only a small fiscal expansion is required to push a weak economy out of recession and boost inflation to the central bank’s target, while also achieving a sizable reduction in the government debt-to-GDP ratio. Moreover, because higher expected inflation tends to raise the policy rate, the zero lower bound is not a constraint on implementation. These results are closely aligned with those reported by Gali (2016), which are derived in a similar framework. What accounts for these powerful expansionary effects? To answer this question, it is helpful to express the monetary policy rule – which links growth in the money supply (which consists entirely of currency in our model) to fiscal spending – as a reaction function in terms of the policy rate. We show that the reaction function under a money-financed program can be interpreted as a form of flexible price level targeting in which the central bank’s policy rate responds to real activity (consumption in our model), as well as to the gap between the price level and its target level. The key twist is that the targeted price level path shifts upward in response to the rise in government spending, with the magnitude of the shift depending on the size and persistence of the government spending hike. Under a pure price-level targeting regime, the fiscal expansion would have to be financed with debt and the result would be a peak multiplier of a bit less than unity with relatively littleseignioragerevenue. Bycontrast, underthealternativereactionfunctionwithanupwardshift in the price level target, the central bank allows real interest rates to decline sharply, fueling an output rise that is sufficiently large to push the price level up to its new target path.3 The implied rise in nominal currency, in turn, generates enough seigniorage to finance the fiscal expansion. As we will show below, the change in the monetary policy reaction function is critical for the effectiveness of the money-financed program. Indeed, since we assume that interest is paid on 3Thelong-runeffectsonthepricelevelaredeterminedbythefamiliarrelationshiplinkingrealseignioragerevenue to currency growth scaled by the ratio of currency to GDP (see Fischer (1982)). Because currency is a small share of GDP, a fiscal expansion causes a large rise in the price level. While this implication is likely to be robust, the contour of the inflation response – as well as of those of output and debt – will depend on the particular features of our model. 2

reserves, the central bank could finance a fiscal expansion with purchases of government securities, and commit to maintaining the increased holdings of securities indefinitely, but still use interest on reserves (and perhaps other tools, such as overnight reverse repurchase agreements) to maintain an unchanged interest rate reaction function.4 In this case, interest payments from the central bank would substitute for interest payments from the Treasury. While such an approach would still provide fiscal stimulus financed by the central bank, it would not provide the additional boost to economic activity and inflation that would come from easier current and future monetary policy.5 Because a money-financed program involves a dramatic shift in the central bank’s normal reaction function, achieving near-term output stimulus from such a program is contingent on the public rapidly understanding the new strategy and regarding it as credible. However, it seems likely that it would be difficult in practice to communicate the new strategy to the public because the resulting period of high inflation would likely be seen as at variance with the objective of low and stable inflation that many central banks have adopted in recent years. In this vein, because raisingrevenuethroughseignioragedependsonthebaseoftheinflationtax–andhenceoncurrency demand–theriseinthepricelevelassociatedwithagiven-sizedmoneyfinancedprogramhingeson the specific features of the currency demand function, which could evolve in unexpected ways over time. Accordingly, the central bank would have to explain how the target path for inflation in the medium-run would depend not only on the size of the fiscal stimulus program(s) authorized by the government,butalsoonpossiblechangesinthepublic’sdemandforcurrency. Asecondissueisthat – even if the public did understand what the new policy entailed – the public would likely doubt that the central bank would carry through with a strategy that permitted a large and persistent rise in inflation. In particular, a money-financed program could suffer from a “time-inconsistency” 4 Ireland (2014) shows that with the payment of interest on reserves, the central bank can choose the level of reserve balances essentially independently of its monetary policy stance. 5 Someauthors(e.g.,CechettiandSchoenholtz(2016)andBorioetal(2016))haveinterpretedamoney-financed program more narrowly along these lines. 3

problem, as policymakers could well have a strong incentive to renege on their promise to allow inflation to remain high once the economy had recovered. For both of these reasons, the shift in monetary policy might not do much to amplify the effects of the fiscal stimulus on output; and inflation might not respond for a long time until the public became convinced of the authorities’ resolve to follow through with the new policy.6 Accordingly, we modify our model to allow for the possibility that agents only gradually come to perceive a shift in the central bank’s reaction function following the announcement of a moneyfinanced fiscal program. Specifically, we adopt a Kalman filtering framework in which agents are uncertainwhethertheincreasesinthemoneysupplyassociatedwithanannouncedmoneyfinanced program are only temporary – and thus will be reversed – or whether they reflect permanent increases in money as called for by the program. We show that if the learning process is gradual, the boost to output and inflation may be small or negligible in the near-term, and only become substantial at longer horizons; an implication is that a money-financed program may be much less effective in providing stimulus in a recession than implied by the baseline model. The boost to output may be small even at longer horizons if agents react to the rise in inflation by adjusting prices more frequently, in which case the main effect of a money-financed program would be a large rise in inflation. Giventhelikelycommunicationsandcredibilityissuesraisedbymoney-financedfiscalprograms, we then consider more limited forms of monetary and fiscal policy coordination that involve more modest shifts in the monetary policy reaction function that might be more credible and easier to communicate. For example, we consider the possibility that the central bank would commit to a reaction function that puts a larger weight on inflation and output gaps than under the baseline rule, implying a more accommodative response to the fiscal stimulus in a deep recession.7 Such 6 Such a major change in the monetary policy framework could even cause output to fall due to increased uncertainty and an associated widening of risk premiums. See Bernanke (2010). 7 Nominal GDP targeting may have broadly similar implications. For a discussion of the possible benefits of 4

a commitment by the central bank would tend to make fiscal expansion more attractive, since it would boost the output effects of any given-sized fiscal program, while reducing the upward pressure on the debt-to-GDP ratio compared with the baseline policy. While these implications are qualitatively similar to a money-financed program, we see an incremental shift in monetary policy as more likely to be credible, in part because it poses much less risk of creating outsized and costly increasesininflation.8 Moreover, especiallyindifficulteconomiccircumstancesinwhichthecentral bankcannotsignalachangeinthereactionfunctionbyloweringthepolicyrate,acoordinatedfiscal program can boost the public’s confidence in policymakers’ ability to achieve their objectives.9 We then briefly consider some historical examples of monetary-fiscal cooperation to illustrate some of our model results. These examples show that money-financed fiscal programs can be very effective in boosting output and inflation. For example, the U.S. experience during and after the Civil War and the French experience during and after World War I provide clear evidence of the potentially substantial economic effects of such policies. Our examples also suggest that the effectiveness of these policies depends on the clarity and credibility with which they can be communicatedtothepublic. Inparticular,wenotethatsuchpoliciesmaybeeasiertocommunicate ifthecountryisonthegoldstandard,sothatachangeinthegoldparitycanbeusedtocommunicate the change in policy regime, and also if the change in policy comes about in response to a rare and visible challenge to the country. Such a challenge may be posed by a large war, of course, but can also, as in the case of Japan in the 1930s, reflect the effects of a substantial economic depression.10 We also note that without the gold standard, the central bank may find it difficult to communicate a higher desired price level without losing control of inflation expectations. For nominal GDP targeting coordinated with expansionary fiscal policy, see the comments by Woodford in Reichlin, Turner, and Woodford (2013). 8 As the price level only increases modestly, only a small fraction of the fiscal expansion is ultimately financed through seigniorage. 9 It bears emphasizing that even these more limited forms of monetary-fiscal coordination may pose important risks,includingtocentralbankindependence(asdiscussedbelow). Accordingly,centralbanksmaywellopttoprovide additional monetary stimulus without explicit coordination with the fiscal authority. 10 See Romer (2014) for a discussion along these lines of U.S. monetary policy in the Great Depression. 5

example, in the United Kingdom in the 1960s expansionary monetary and fiscal policy succeeded in boosting economic activity and inflation, but the result was a sustained period of much higher inflationthatproveddifficult,andultimatelycostly,tobringbackdown. Conversely,acentralbank that has built a reputation for low and stable inflation may find it difficult to credibly promise even the temporary rise in inflation that would be implied by a period of monetary-fiscal cooperation. We see the challenges that the Bank of Japan has faced in boosting inflation expectations in recent years in this light. This paper is organized as follows: Section 2 presents the benchmark model and calibration, while Section 3 discusses our simulation results. Section 4 considers monetary-fiscal cooperation from a historical perspective, while Section 5 concludes. 2. Benchmark Model Our benchmark model builds on the workhorse New Keynesian model by incorporating a demand function for money (which consists exclusively of currency). In the next section, we show how a rule for setting currency growth to yield a given amount of seigniorage can be re-interpreted in terms of an instrument rule for the policy rate. 2.1. Households Householdsderiveutilityfromconsumption, c ,andfromtheirholdingsofrealmoneybalances, m , t t but experience disutility from the hours they spending working, n . The period utility function of t households is assumed to take the separable form: u(c ,n ,m ) = [c t −ψ c ca t−1 +ν 0 ξ t ]1− σ 1 −1 −χ n1 t +χ − µ 0 [max{v∗− m t ,0}]1+µ (1) t t t 1− 1 01+χ 1+µ t ca σ t where the discount factor β ∈ (0,1), σ ≥ 0, 0 ≤ ψ < 1, χ ≥ 0, χ > 0, µ > 0 , and µ > 0. c 0 0 The subutility function over consumption allows for external habit persistence in consumption, 6

with ca denoting lagged aggregate consumption, which households take as given in deciding on t−1 their own consumption. The consumption taste shock ξ is an exogenous AR(1) process with a t persistence parameter of ρ . The final term of equation (1) implies that real balances – expressed ξ as a ratio to aggregate consumption – are valued at the margin until reaching a stochastic bliss point of v∗. For simplicity, we assume that the scaling factor is aggregate consumption, ca, which t t is taken as given by the household; this formulation implies that the consumption Euler equation doesn’t depend on the level of real balances, consistent with most empirical analysis. Households face a budget constraint that may be expressed in real terms as: 1 b +m +c = (1−τ )w n + ((1+i )b +m )+d −τ (2) t t t n,t t t t−1 t−1 t−1 t t 1+π t The household’s after-tax income consists of labor income, (1−τ )w n (where w denotes the n,t t t t real wage and τ is a distortionary labor income tax), dividends, d , from the ownership of firms, n,t t minusthelump-sumtaxes, τ , paidtothegovernment. Thehouseholdusesthisincometopurchase t consumption goods, increment its real bond holdings, b = Bt (nominal holdings B divided by the t Pt t price level P ), and accumulate real balances, m = Mt. The real return on bonds is 1+it−1, where t t Pt 1+πt i is the nominal interest rate and π = Pt −1 is the inflation rate; the gross real return on t−1 t Pt−1 money is simply (1+π )−1. t The household’s problem consists of choosing its consumption, labor hours, real balances, and bond holdings to maximize its discounted utility E (cid:80)∞ βt U(c ,n ,m ) subject to the budget 0 t=0 t t t constraint given by equation (2). The first order conditions imply the usual consumption Euler equation and household labor supply decision, respectively: (cid:26) (cid:27) 1+i t λ = βE λ (3) t t t+1 1+π t+1 u − nt = w (1−τ ), (4) t n,t λ t 7

where λ is the marginal utility of consumption (which depends partly on lagged consumption). t Household money demand can be expressed as: (cid:18) (cid:19) m 1 µ (v∗− t )µ = λ c 1− , if i > 0 (5) 0 t c t t 1+i t t t Given the opportunity cost of holding money balances when the (net) interest rate is positive, real money demand (expressed relative to consumption) is less than its satiation level v∗. As in t Eggertsson and Woodford (2003), the money demand function is continuous at i = 0 with mt ≥ v∗ t ct t if i = 0. This formulation is useful because it allows us substantial flexibility to calibrate the t interest semi-elasticity of money demand, as can be seen from the log-linearized form that we work with below: v∗ m (cid:101)t = (cid:101) c t −φ c (λ(cid:101)t + (cid:101) c t )−φ i i(cid:101)t +( v )v t ∗ , if i t > 0 (6) where φ = v v ∗ −1 and φ = v v ∗ −1 1 . Under log utility over consumption and abstracting from c µ i µ i(1+i) habit, real money balances vary directly with consumption with a unit coefficient (since λ(cid:101)t = v∗ −1 −c ), and the semi-interest elasticity of money demand varies directly with v . (cid:101)t µ 2.2. Firms and Price-Setting We assume that there is a single final domestic output good Y that is produced from a contint uum of differentiated intermediate goods Y (f) according to the Dixit-Stiglitz technology Y = t t (cid:104) (cid:82)1 1 (cid:105)1+θp 0 Y t (f)1+θp df where θ p > 0. Firms that produce the final output good are perfectly competitive in both product and factor markets, purchasing intermediate goods, Y (f), at prices P (f) t t to minimize the cost of producing Y . The demand schedule for each intermediate good derived t from this cost-minimization problem is of the form: (cid:18) (cid:19)−(1+θp) P t (f) θp Y (f) = Y . (7) t t P t (cid:20) (cid:82)1 −1 (cid:21)−θp where P t is the aggregate price index (i.e., P t = 0 P t (f)θp df ). 8

Intermediate good f is produced by a monopolistically competitive firm according to a Cobb- Douglas production function: Y (f) = K (f)α(Z L (f))1−α, (8) t t t t where Z denotes a stationary shock to the aggregate level of productivity. The intermediate goods t producers face perfectly competitive factor markets for hiring capital, K(f), (at a real rental price of r ) and labor, L (f), (at a real wage of Wt). The first-order conditions for the cost-minimizing Kt t Pt input choices imply that all intermediate producers have an identical real marginal cost per unit of output MCt. Real marginal cost can be expressed as the ratio of the real wage to the marginal Pt product of labor: MC t = W Pt t = W Pt tL t , (9) P MPL (1−α)K t t noting that the aggregate ratio of labor to capital appears in (9) because all firms choose the same factor proportions, and that the aggregate capital stock is assumed fixed. Intermediate-goods-producingfirmssetpricesaccordingtoCalvo-stylestaggeredcontracts,with firm f facing a constant probability, 1−ξ , of being able to re-optimize its price, P (f). Each firm p t that is allowed to reoptimize chooses its price (Popt(f)) to maximize: t ∞ max E (cid:88) ξjψ (cid:104) (1+π)jPopt(i)−MC (cid:105) Y (i), (10) t p t,t+j t t+j t+j Popt(f) t j=0 subjecttoitsdemandcurve(7), whereψ isthestochasticdiscountfactor(i.e. βjE λt+j)). The t,t+j t λt term (1+π)j reflects that the price of firms not able to reset their price is assumed to increase at the steady-state inflation rate of π each period. The first order condition (together with the identity for the evolution of the aggregate price index) yields the standard New Keynesian Phillips Curve. 9

2.3. Fiscal Policy It is convenient to consider the fiscal authority (Treasury) and central bank as a single consolidated entity (the “government”). The government finances its purchases of real goods and services, g , t either through taxation, real seignorage revenue, or by issuing debt. Accordingly, the government’s flow budget constraint, which determines the evolution of government debt, b , is: t (cid:18) (cid:19) 1+i t−1 b = b +g −τ −τ w n −s , (11) t t−1 t t n,t t t t 1+π t where s is real seignorage revenue, and recalling that τ is a lump-sum tax on households and τ t t n,t the labor income tax rate. Real seignorage revenue in turn is equal to the new nominal monetary liabilities issued by the government divided by the price level (i.e., the purchasing power of newly issued money): M −M m t t−1 t−1 s = = m − (12) t t P 1+π t t We assume that government spending follows an exogenous AR(1) process of the form: g = (1−ρ )g +ε . (13) t g t−1 gt andthattheincometaxrateτ isconstant. Thus,lump-sumtaxesadjusttosatisfythegovernment n budget constraint (11). 2.4. Market Clearing The aggregate production function is given by: y = a kan1−α (14) t (cid:101)t t where a denotes the level of technology scaled by price dispersion (see Woodford (2003)), and (cid:101)t (cid:82)1 (cid:82)1 aggregatelaborandcapitalareL = L (f)df andK = K (f)dfK,respectively. Theaggregate t 0 t 0 t resource constraint is: y = c +g (15) t t t 10

2.5. Monetary Policy Our benchmark assumption is that monetary policy sets seigniorage revenue to be proportional to the shock to government spending on goods and services, i.e., to the deviation of government spending from its steady state level. M −M M −M M (cid:18) g −g (cid:19)φ t t−1 t t−1 t−1 t s = = = (16) t P M P g t t−1 t We devote considerable attention to the case in which any shock to government spending is fully financed by seigniorage (i.e., φ = 1). However, we also consider cases in which government spending is only partially financed by seigniorage. As we will discuss, this money-based rule can be recast in terms of a reaction function for the policy rate that makes it easier to interpret. 2.6. Calibration We calibrate our model at a quarterly frequency using fairly standard parameter values. The discount factor of β = 0.995 implies a steady state real interest rate of 2 percent (at an annualized rate). With a steady state inflation rate of 0 percent (i.e., π = 0), the steady state nominal interest rate is also 2 percent. We set the intertemporal substitution elasticity σ = 0.5, so that the interest elasticity of aggregate demand is somewhat lower than under log utility.11 The habit parameter ψ c is set to 0.8. This value is on the higher side of the range of estimates in the empirical literature, but helps our model generate a fairly plausible path for the government spending multiplier, even if somewhat lower than estimated by e.g., Blanchard and Perotti (2002). The Frisch elasticity of labor supply of 1 = 0.5 and capital share of α = 0.3 are in the typical range specified in the χ literature. The government share of steady state output is set to 20 percent (g = 0.2), which is y close to the average federal government spending share in U.S. GDP. 11 Thescaleparameterontheconsumptiontasteshockν issetto0.001(thisparameterissettohaveanegligible impact on the slope of the IS curve). 11

The responsiveness of inflation to marginal cost plays a key role in determining how monetary policy actions affect output and inflation. We assume ξ = 0.95, so that the mean duration of p price contracts is 20 quarters – implying a very flat Phillips Curve slope that seems consistent with empirical evidence for the United States, at least in recent years. The specification of money demand implies a unitary long-run elasticity with respect to consumption. We impose a short-run interest rate semi-elasticity of money demand equal to 2.5 (when expressed at an annual rate). Given that the share of real balances in consumption is set to 0.4, this implies that a 1 percentage point rise in the (annualized) nominal interest rate would reduce the share of currency to GDP by 1 percentage point.12 We discuss the monetary policy rule in more detail below. Finally, we solve the log-linearized model using the AIM algorithm. 3. Effects of a Money-Financed Fiscal Program We begin by using our model to show how a fairly small money-financed fiscal program could – at leastinprinciple–providealargeboosttooutputwhilealsoreducingthegovernmentdebttoGDP ratio. Thus, as seen in Figure 1, we consider a rise in government spending that equals only 0.4 percent of GDP in the initial quarter, and which cumulates to 1 percent of GDP; the persistence of the shock is 0.9. Both the design of the scenario and our results are close to those reported in Gali (2016). The red dashed lines show the effects of the government spending hike under the assumption that monetary policy follows the Taylor (1993) rule. Although output and inflation both rise, the government spending multiplier (not shown) is less than unity, reflecting that the central bank raises both nominal and real interest rates, which crowds out private demand. The spending multiplier averages about 0.8 percent in the first year following the shock, which is closely in line 12 The 40 percent figure reflects quarterly consumption. Currency is roughly 10 percent of annual consumption spending. 12

with most empirical estimates of the spending multiplier based on U.S. data prior to the financial crisis (Ramey and Zubairy, forthcoming). The spending hike is almost exclusively debt-financed. The ratio of government debt to GDP rises because the higher government spending generates persistent primary deficits, and because the higher real interest rates raise the cost of financing existing debt. Seigniorage actually falls in the short-run as higher nominal interest rates reduce real money demand, though the eventual rise in the price level translates into a small amount of seigniorage at longer horizons. Policiesthatboostdebtmayberegardedasundesirableorunsustainable,especiallyineconomies with high levels of public debt. Accordingly, our second scenario considers a money-financed fiscal program. In this simulation we assume the same fiscal expansion as before, but also assume a radical shift in monetary policy, under which the central bank commits to raising the demand for nominal currency permanently by the amount required to finance the increase in government spending. (The increase must be in currency, since we assume that interest is paid on reserve balances.) In order to boost currency demand by the required amount, the price level must eventually rise by enough that the public voluntarily demands this additional nominal currency. This change in monetary policy is assumed to be fully understood by the public and completely credible – an assumption to which we will return below. The blue solid lines in Figure 1 show the results of this money-financed fiscal program in our model. The money-financed program generates a large and persistent rise in the inflation rate. The inflation rate rises by more than 2 percent above baseline, and eventually causes the price level to rise by 10 percent. The large rise in the price level in response to this fairly small increase in government spending reflects that the inflation tax base – the ratio of money balances to nominal GDP – is small, so that the money stock must rise by a substantial amount in percentage terms. Becauserealinterestratesdeclinemarkedly,GDPexpandsbynearly11 percentrelativetobaseline, 2 13

consistent with an implied spending multiplier of around 5. The money-financed program elicits a largefallinthedebt-to-GDPratioofaround13percentagepointsafteradecade. Whileseigniorage revenue rises by the size of the government spending hike by design, most of the improvement in the debt-to-GDP ratio reflects a rise in tax revenues due to the output expansion, and also the depressing effects of low real interest rates on interest expenses and hence outstanding debt. Resultsalongtheselineshaveledanumberofeconomiststoconcludethatcentralbank-Treasury coordinationtomonetizefiscalspendingcouldbeapotenttooltoconfrontcyclicaldownturns: Only a small dose of fiscal spending is required to provide a powerful boost to GDP and inflation, and such a program would have the collateral benefit of reducing the government debt burden. Support for including monetization in the arsenal of central bank tools has undoubtedly been fueled by the long period in which many central banks have been constrained by the effective lower bound (ELB) on policy rates since the financial crisis. An attractive feature of a money-financed fiscal program in a liquidity trap is that the rise in inflation expectations and aggregate demand it engenders tends to boost the nominal interest rate, and hence can help “lift” central banks off the ELB more quickly. Figure 2 shows the effects of the same (small) rise in government spending against the backdrop of a deep recession that would pin the policy rate at zero for three years absent the fiscal stimulus. In contrast to our previous results in Figure 1, which are reported in deviations from the steady state baseline, Figure 2 shows the responses in levels, with the baseline generated by a transitory preference shock that increases desired savings. The boost to output and inflation from fiscal stimulusunderastandardmonetarypolicyreactionfunction–thepromisetocontinuefollowingthe TaylorruleafterliftofffromtheELB–isbarelynoticeable. Thespendingmultiplierisconsiderably larger than in normal times because real interest rates fall, which crowds in private demand (c.f., Christiano, Eichenbaum, and Rebelo 2011, and Woodford 2011); even so, a much larger fiscal 14

program would be required to close output and inflation gaps.13 By contrast, the money-financed program quickly erases most of the output gap, moves inflation above target within two years, and induces some decline in government debt relative to GDP (rather than a sizeable runup). Notably, the nominal interest rate rises from the ELB shortly after the shock, and attains its long-run value within a couple of years. 3.1. Re-interpreting the central bank’s reaction function While the central bank’s reaction function under a money-financed program is typically thought of as a money supply rule, it is helpful to express the rule in a more familiar form in terms of the policy rate. In general, the policy reaction function implied by monetary financing depends on the currencydemandfunction. Underthereasonableassumptionsaboutcurrencydemandinourmodel —that real currency demand varies directly with activity, and inversely with the policy rate—the central bank’s reaction function under a money-financed fiscal program can be interpreted as a form of “flexible price-level targeting,” in which the policy rate varies with the gap between the price level and its target path, and also with real activity (output or consumption). The crucial twist is that the target price level path varies with the size and persistence of the fiscal expansion, rather than remaining constant or growing at fixed rate (e.g., 2 percent per year). In this vein, the interest rate reaction function can be derived by noting that the log-linearized form of the monetary policy rule in equation (12) can be expressed: m (∆m +π ) = g g (17) y (cid:101)t (cid:101)t y(cid:101)t where m is the ratio of real balances to steady-state output, and g is the steady-state government y y expenditure share of output.. Thus, a money-financed increase in government spending requires some combination of an expansion in real money demand (∆m ) and higher inflation. Substituting (cid:101)t 13 Note that Figure 2 shows the output gap (to keep all responses in levels). The output response is somewhat larger given that higher government spending boosts potential GDP. 15

the log-linearized money demand function (6) into equation (17) and solving for the policy rate yields: 1 (cid:18) g v∗ (cid:19) ∆(cid:101)i t = φ π (cid:101)t − m y g (cid:101)t +φ c (∆ (cid:101) c t −( v )∆v t ∗) , if i t > 0 (18) i y where we have abstracted from habit persistence for expositional simplicity. This “super-inertial” reaction function can be written alternatively in terms of the level of the policy rate as: 1 (cid:18) v∗ (cid:19) (cid:101)i t = φ p (cid:101)t −p (cid:101) ∗ t +φ c ( (cid:101) c t −( v )v t ∗) , if i t > 0 (19) i where the price level target, p∗, evolves according to: (cid:101)t g p∗ = p∗ + y g (20) (cid:101)t (cid:101)t−1 m (cid:101)t y Abstracting from p∗, the interest rate reaction function (19) can be interpreted as consistent (cid:101)t with a form of flexible price level targeting, with consumption rather than output the relevant activity measure. The key difference is that a money-financed fiscal expansion raises p∗, which (cid:101)t can be regarded as tantamount to a time-varying price level target that varies with the size and persistence of the government spending expansion. Thisframeworkprovidesausefulwaytohelpunderstandwhyamoney-financedfiscalexpansion generates a much larger rise in output and inflation than under the Taylor rule. Figure 3 provides a decomposition along these lines. In particular, this figure reports exactly the same responses to a government spending shock as in Figure 1 under both the Taylor rule (dashed red) and moneyfinancedprogram (solidblue). However, Figure 3also showsthe effects ofthe government spending hike in the case that monetary policy adopted flexible price-level targeting – consistent with the reaction function in equation (19) – but leaving the price level target path p∗ unchanged (the green (cid:101)t dash-dotted lines). The striking feature of the figure is that the shift to flexible price level targeting alone induces only a slightly bigger output response (the multiplier in the first year is only about 10 percent larger). Thus, the vastly larger multiplier under the money-financed program occurs 16

because agents regard it as implying a large upward shift in the price level target path over time (the black dotted line in the upper right panel shows the target path p∗). (cid:101)t Giventhatagentsimmediatelyrecognizethisshiftinthetargetpricelevelpath,itisunsurprising that there are large expansionary effects on output. The central bank must lower the trajectory for the real interest rate in order to raise output by enough to boost the price level toward this new target path, and hence generate the seigniorage needed to finance the fiscal expansion. Figure 3 illustrates how most of the decline in the real interest rate under a money-financed program is due to the upward shift in the price-level target path (the difference between the solid blue and green dash-dotted lines in the middle left panel). The strong expansion of output and inflation causes the nominal interest rate to rise (as was seen in Figure 2). 3.1.1. Role of Money Demand Equation (19) underscores how the effects on output of a money-financed fiscal expansion depend critically on the specific features of money demand. The price level target path would rise more than in Figure 3 if real money demand was a smaller share of output – as the base for the inflation tax would be smaller – and conversely, rise more if money demand was larger. Thus, as seen in Figure 4, while our baseline calibration of money demand implies that the price level rises about 10 percent relative to baseline under the benchmark case, the price level would rise 20 percent under a calibration in which money demand were only half as large, and the effects on output and interest rates would be commensurately larger. 3.1.2. Monetary Financing and the Phillips Curve Slope The slope of the Phillips Curve plays an important role in determining how the rise in nominal money demand associated with a money-financed fiscal program is distributed between output and 17

inflation. Under our benchmark calibration with a low Phillips Curve slope, a large and relatively persistent output expansion is required in order to boost the price level to its new target path. However, if inflation was more responsive to output, a smaller and more transient output expansion could achieve the same rise in nominal demand, so that a given-sized money-financed fiscal expansion would imply less GDP stimulus. The red dashed lines in Figure 5 illustrate the implications of the same rise in government spending under a steeper Phillips Curve slope corresponding to a Calvo parameter of ξ = 0.8, rather than ξ = 0.95 as in our benchmark calibration. Output p p peaks at a little over 1 percent above baseline under this alternative calibration, while the peak rise in inflation is nearly twice as large. The shorter duration of the output expansion translates into a less persistent improvement in the primary budget balance, and, accordingly, the government debt-to-GDP ratio also declines by less. As we will discuss more below, when considering historical episodes, the prospect of a relatively muted inflation response to a money-financed program is likely to be greater when the central bank isperceivedasallowingaone-timeshiftinitspriceleveltarget,perhapsinresponsetoextraordinary circumstances such as a major war or depression. Conversely, repeated efforts to monetize fiscal deficits would seem likely to engender a much larger inflation response – and correspondingly, less persistent effects on real interest rates and GDP – as agents would likely react by re-setting prices more frequently, and possibly by adjusting upward their views about longer-term inflation. 3.2. The Effects of Imperfect Credibility Given that monetary financing involves a dramatic shift in the policy reaction function, it seems likely that the private sector would either fail to understand what the new policy regime entailed, or regard the central bank’s commitment to the new regime as somewhat tenuous. Accordingly, we next try to capture some of this uncertainty by assuming that agents must solve a signal extraction 18

problem: They see the central bank printing money, but must decide whether the money stock hike presages monetization of the new government spending, or will be reversed.14 Tosetupthesignalextractionproblemformally, webeginbyrecallingthatunderfullmonetary financing, money growth is determined according to: g y M = M + g (21) t t−1 (cid:101)t m y We now consider the possibility that only some component of the new government spending gmf (cid:101)t is financed by money creation, while the remainder is financed by issuing debt gdf (thus, g = (cid:101)t (cid:101)t gmf +gdf). Moreover, we assume that the money stock may also be buffeted by other shocks, so (cid:101)t (cid:101)t that money growth is given by: g M = M + y (gmf +∆e ) (22) t t−1 m (cid:101)t Tt y Themoney-financedcomponentgmf isitselflinkedtoanunderlyingshocke viathelinearrelation (cid:101)t gt gmf = ψu where u has the same persistence as the actual government spending shock g : (cid:101)t gt gt t u = (1−ρ )u +e (23) gt G gt−1 gt where the innovations e and e are assumed to be N(0,1), and uncorrelated with the innovation gt Tt to government spending ε in expression (13). gt This framework provides a stylized way of analyzing how a money-financed program might play outunderdifferentassumptionsabouttheinformationavailabletoagents. Broadlyspeaking,agents wouldtypicallyexpectfiscalspendingtobedebt-financed: Thisiscapturedbyourassumptionsthat the money-finance innovation e is uncorrelated with the fiscal innovation, and that the scaling gt parameter ψ is very small. Hence if the authorities did opt to finance the higher government 14 Inoursimplemodel,areversalwouldrequirethecentralbanktoscalebackcurrencyincirculationandsubstitute interest-bearing debt. In practice the central bank could finance the fiscal expansion initially by expanding the monetary base – and thus depart temporarily from its normal reaction function – but later pay interest on those reserves (consistent with a return to its usual reaction function). 19

spending by printing money, this would be reflected in an extremely large rise in u (given that gt ψ is small). Under full information, agents would observe the shock u directly, and accordingly, gt immediately change their views about the long-run money stock and price level, recognizing that the money shock would have to rise enough to finance the new government spending. By contrast, underimperfectinformation,weassumethatagentsonlyseetheactualmoneystockM , butcannot t distinguish the money-financed component u from random variation in the money stock that is gt captured by e . Hence, they must solve a signal extraction problem to infer these components Tt based on their incoming observations about M . Specifically, they use the Kalman filter with the t observation equation given by:   M t M t = H(cid:48)z t = (cid:104) 1 0 0 (cid:105)  u gt   , (24)   e Tt and where the underlying state vector z in turn is perceived to follow a first-order vector autoret gression:        M t 1 ψ(1−ρ G ) −1 M t−1 ψ 1 (cid:34) (cid:35) e        gt  u  =  0 1−ρ 0  u + 1 0  (25)  gt   G  gt−1    e Tt e 0 0 0 e 0 1 Tt Tt−1 Figure 6 shows the implications of imperfect credibility for output, inflation, and government debt. Whileagentsareassumedtohaveacompleteunderstandingofhowgovernmentspendingwill evolve(itisthesameshockasinFigure1),theybelieveinitiallythatitwillbelargelydebt-financed. The speed at which agents update their views about the state – and critically, about the persistent shock u that determines the extent of monetary financing – depends on the Kalman gain, which gt in turn depends on the size of the parameter ψ. Specifically, the parameters of the Kalman filter are set so that agents initially believe that only a small fraction of the increase in spending will be monetized, and accordingly, initially expect most of the rise in money to be reversed. However, as agents are surprised by ongoing increases in the money stock (middle right panel), they come to expect that more of the government spending hike will be monetized. 20

OutputandinflationrisemuchmoregraduallyinthissimulationthaninFigure1,withthepeak impact on output and inflation occurring 6-7 years after the policy is implemented. Thus, to the extent that the policy aims to support the economy in a recession, the stimulus is less “well-timed” than under the extreme assumption of full information. 3.3. More Limited Forms of Monetary-Fiscal Cooperation A money-financed fiscal expansion may be appealing in some circumstances because it has the potential to raise nominal aggregate demand while reducing the debt-to-GDP ratio, perhaps significantly. Nevertheless, our analysis suggests several major drawbacks of such an approach. First, the effects on inflation are likely to be very large given plausible estimates of the inflation tax base, even for a modest-sized fiscal expansion. Second, the size and timing of the effects on inflation are difficult to predict, and correspondingly, would compound the difficulties of explaining such a program to the public. Not only would such a program involve a major shift in the central bank’s reaction function – towards flexible price level targeting – but the path of the price level target would have to be adjusted based on unexpected changes in money demand. Finally, while we have assumed in our model simulations that long-run inflation expectations remained well anchored at the central bank’s target, this assumption may seem unrealistic given the large and persistent rise in inflation above target following a money-financed fiscal program. Inflation would rise even more than in our simulations if high realized inflation pushed up long-run inflation expectations, and the central bank might have to take costly actions later on to re-anchor expectations. Given these considerations, in this section we explore the potential benefits of complementing fiscal stimulus with a more modest shift in the monetary policy reaction function. Such a shift would be “in the spirit” of a money-financed program insofar as it would be intended to make fiscal expansion more effective in raising output, while at the same time putting less upward pressure on 21

public debt than under the baseline rule. In this vein, Figure 7 considers a deep recession baseline (the black solid lines) and government spending expansion against that backdrop. We contrast a government spending expansion under the standard Taylor rule (the red dashed lines) with an alternative in which the central bank commits to following a Taylor rule that puts a weight of unity on the output gap, rather than 0.5 as under the standard Taylor rule, and also puts a larger weight on deviations of inflation from its target. The government expansion equals 1 percent of GDP for 10 quarters at an annualized rate, so that it cumulates to 2.5 percent of GDP rather than the 1 percent of GDP considered in Figure 1; our assumption of a larger fiscal package reflects that fiscal spending is clearly less potent than under monetization with full credibility, so that more is needed to alleviate a recession. But the key point is that the more aggressive rule causes output and inflation to rise more in response to the fiscal expansion, and the debt-to-GDP ratio rises by considerably less than under the standard Taylor rule.15 Figure 8 examines the consequences of fiscal stimulus coupled with a more substantial shift in the policy framework to flexible price level targeting (again against the same recession baseline shown by the black solid lines). The key difference relative to the monetary financing case is that the central bank does not commit to adjust the price level target path to ensure that the fiscal expansion is fully monetized. Thus, the first alternative – shown by the red dashed lines – involves a shift toward flexible price-level targeting in which the target path is unchanged. Given theadmittedlystrongassumptionthatthenewpolicyruleisimmediatelyunderstoodbythepublic and regarded as fully credible, the combined monetary and fiscal program is effective in keeping inflation close to target and in eliminating most resource slack. Moreover, the resulting stronger nominal aggregate demand causes the debt-to-GDP ratio to fall relative to the baseline (although 15 Relatedly, Engen, Laubach, and Reifschneider (2015) show that the effects of asset purchases are enhanced by the adoption of a more accommodative monetary policy reaction function. 22

only a small fraction of this decline is due to increased seigniorage). In a second alternative, the target path for the price level is allowed to grow at an average rate of about 1 percent per year faster than the baseline for several years, with the price level target risingby5percentintheverylong-run. Thispolicyprovidesmorestimulustoinflationandoutput, and thus results in a larger decline in the debt-to-GDP ratio. Given the greater rise in the price level, this policy rule also yields noticeably larger seigniorage revenue than in the alternative with a constant price level target, and thus moves further in the direction of monetary financing. Even so, a key difference – including in communicating the policy – is that the target path for the price level would not be linked explicitly to raising a specific amount of seigniorage, and hence would not be required to shift in response to unexpected changes in money demand. Of course, the downside of this approach is that it leaves the debt-to-GDP ratio at a significantly higher level than in the case of full monetary finance, although the ratio still declines as a result of the program. 4. Monetary-Fiscal Cooperation: Some Historical Examples Although the importance of central bank independence for the achievement of low and stable inflation has been emphasized by monetary economists in recent decades, monetary-fiscal cooperation similar to that discussed in this paper has been common in the past. In particular, there are many historical examples of countries that have used monetary policy to help finance fiscal actions in wartime.16 In effect, such cooperation allowed for the financing of war expenditures with seigniorage. In many cases, this approach required a temporary suspension of the gold standard. For example, U.S. government during the Civil War issued unbacked paper money – the “greenbacks” – to finance war expenditures that totaled about 10 percent of pre-war GDP.17 In addition, under the National Banking Act, newly-chartered national banks issued national bank notes that were 16 See Rockoff (2014) for a summary of this literature and a discussion of U.S. examples. 17 Rockoff (2014) reports that total issuance of greenbacks reached nearly $500 million, while Historical Statistics of the United States reports nominal GDP at the start of the war was less than $5 billion (table Ca9-19). 23

backed only by government securities, an implicit monetization of the debt. In total, the nominal money supply increased by about 17 percent of pre-war nominal GDP. Over the same period, federal debt increased by about $2.6 billion, or nearly 60 percent of pre-war GDP.18 The result of the fiscal and monetary stimulus was a rise in output, as well as an increase in the general level of prices of more than three quarters over the course of the war.19 However, unlike the situation modeled above, the United States ultimately returned to the gold standard at the pre-war parity in 1879, implying that the increase in the price level was temporary, rather than permanent. By contrast, French governments after the First World War ultimately allowed the value of the franc to decline by 80 percent before it was de facto repegged to gold in 1926. As discussed in Bordo and Hautcoeur (2003), France had not increased taxes to pay for the war, but had rather depended on borrowing and substantial money growth (which exceeded 60 percent for a time). As a result, at the end of the war, France faced very large challenges, including a debt-to-GDP ratio of more than 180 percent, large ongoing budget deficits, prices that had more than doubled relative to pre-war levels, and a large monetary overhang (which they estimate at 17 percent of GDP). These challenges were reinforced by the war-related decline in GDP, which was about 10 percent below its 1910 level in 1919. Given this difficult situation, Bordo and Hautcoeur (2003) note that a return to the pre-war parity would have been extremely difficult, although they argue that had political agreement on burden sharing been reached earlier, the ultimate devaluation in terms of gold could have been limited to roughly 50 percent. Even in that case, however, the result of the monetary financing of a significant portion of the war effort would have been a permanent doubling of the price level. As it was, prices rose by roughly a factor of five over twelve years.20 18 The federal debt figures are taken from Historical Statistics of the United States, Table Ea584–587. 19 Output and price data are from Historical Statistics of the United States, Table Ca9-19. Output is available only for the United States as a whole. This measure increased by about 20 percent over the course of the war, with much of the increase presumably in the North. 20 The permanent increase in prices in France may have helped to support the recovery in output after the war, whichreboundedtomorethan120percentofits1910levelby1925despitethewartimedamage. Bycontrast,inthe UnitedKingdom,thereturntotheprewarparitywasaccompaniedbyweakergrowth,withU.K.outputin1925less than 110 percent of its 1910 level (Bordo and Hautcoeur, 2003, figure 8). 24

An intermediate case arose in the United States at the time of the Second World War, when the Federal Reserve cooperated with the Treasury to help with the financing of the war effort. To do so, the Federal Reserve essentially fixed the yields on Treasury securities at low levels, which also limited price volatility that could have adversely affected financial intermediaries. As noted by Chaurushiya and Kuttner (2003), the Federal Reserve purchased a very large fraction of outstanding Treasury bills for a time in order to keep their yields from exceeding 3/8 percent. By contrast, Federal Reserve purchases of longer-term Treasury securities were modest until after the war, when the ceiling on bill rates was increased. By the end of 1945, high-powered money (currency plus banks’ reserve balances at the Federal Reserve) had roughly doubled from its prewar level to approximately $40 billion, while gross federal debt surged from $51 billion in 1940 to more than $260 billion in 1945, reflecting deficits of roughly $50 billion a year from 1943-45.21 In the face of this extraordinary monetary and fiscal policy, nominal GDP more than doubled, from $101 billion to $223 billion between 1940 and 1945, with real GDP rising more than 70 percent and prices increasing by more than a quarter.22 The inflationary effects of the expansionary war-time policies in the United States were blunted by rationing and price controls until after the war. But the withdrawal of the price controls in 1946 led to a sharp further increase in prices, which rose by another third between 1945 and 1948 – a cumulative increase of roughly two-thirds from pre-war levels. To limit the expansionary effects of its interest rate peg and associated securities purchases, the Federal Reserve greatly increased reserve requirements, with the required reserve ratio peaking at 26 percent in the fall of 1948.23 The recession in 1948-49 eased inflationary pressures until the run-up to the Korean War in 1950. Concerns about inflation at that time sparked discussions between the Federal Reserve and the 21 High-poweredmoneyfiguresaretakenfromthe1940and1946AnnualReportsoftheBoardofGovernorsofthe Federal Reserve System. The Federal Debt figures are taken from Historical Statistics of the United States, Table Ea679–682. 22 Output and price data are from Historical Statistics of the United States, Table Ca9-19. 23 Eichengreen and Garber (1991), pp. 183. 25

Treasury that led, in turn, to the Accord in March 1951. Even following the Accord, the Federal Reserve took steps to limit the increase in interest rates for a time, although large tax increases were also put in place. Eichengreen and Garber (1991) provide an interpretation of monetary policy in the pre-Accord period that is broadly consistent with our description of monetary-fiscal cooperation. They suggest that the Federal Reserve’s efforts to limit the variation in interest rates during this period can be seen as implying a target range for the price level, in which the effects of strong aggregate demand on real interest rates are countered with monetary accommodation so long as the price level is expected to remain within a desired target range. In this view, the low level of longer-term yields that the Federal Reserve supported was feasible because investors expected the price level to fall back to a lower level over time, providing a higher real return on monetary assets. Thus, so long as the Federal Reserve’s price level target was credible, low short-term rates that accommodated the fiscal expansion and allowed prices to increase temporarily were consistent with low nominal longer-term yields. In part this policy mix was effective because investors anticipated that the economy would return to its weak pre-war state after the end of the war. However, with the economy stronger than expected, the Federal Reserve had to tighten policy to maintain the price level target, and the interest rate on bills was increased in the late 1940s, reinforced by the increase in reserve requirements. But as the economy geared up for the Korean war, the Federal Reserve could no longer achieve its price level target without raising interest rates, and the result was the Accord and a move to a tighter monetary policy. Monetary-fiscal cooperation has also been implemented in periods of economic stress outside of wartime. For example, in the early 1930s, Japan recovered relatively early and rapidly from the Great Depression, reflecting the effects of the Takahashi economic policy. This policy, introduced by finance minister Korekiyo Takahashi staring in late 1931, had three parts: a departure from 26

the gold standard and substantial devaluation of the yen; an increase in government spending with explicit financing of government deficits by the Bank of Japan (BOJ); and more accommodative monetary policy, including reductions in the discount rate and an easing of regulations limiting note issuance (see Shizume, 2009). This combination of policies appeared to be quite effective, with Japanese industrial production rebounding strongly in 1932 and wholesale prices advancing at abouta10percentpace, onbalance, in1932and1933afterdecliningevenmorerapidlyin1930and 1931. This strong macroeconomic impact appears to have reflected in part the effect of the clear change in policy regime on inflation expectations, consistent with our earlier modeling results.24 These historical examples make clear that fiscal policy paid for in part with seigniorage revenue can be effective in boosting output and inflation. However, the examples focus on periods of extraordinary fiscal stress caused by wars or depressions, which may help to convince the public that a significant change in policy regime has occurred. That leaves the question of whether such a policy approach could be helpful in less dramatic circumstances. One case along these lines is policy in the United Kingdom in the 1960s. The Bank of England wasnotanindependentcentralbankatthetime,andsothegovernmentcouldimplementmonetaryfiscal cooperation.25 In addition, this policy came at a time when there was no major war and no severe economic crisis, and monetary policy was not constrained by the effective lower bound and inflationwasnotpersistentlylow. Nonetheless,thegovernmentusedcoordinatedpoliciestoprovide accommodation based on the judgment that potential output was higher than it appears to have actuallybeen(NelsonandNikolov2003). Notsurprisingly,theresultwasnotaone-timejumpinthe price level to accommodate the fiscal expansion, but rather a steady rise in inflation over a period of more than a decade. Indeed, even after the price level had increased and there was a devaluation 24Shizume(2009),p. 7. Romer(2014)arguesthattheU.S.departurefromthegoldstandardduringthedepression had similar effects because it marked a credible change in regime. 25 SeeCairncross(1996)foradiscussionofhowtheBankofEngland’slimiteddegreeofoperationalindependence during this period affected its policy choices. 27

in the late 1960s, the government continued to run an accommodative policy, pushing inflation and inflation expectations higher, As a result, there was a prolonged period of undesirably high inflation that was ultimately brought to a close only by a protracted contraction that was required to re-anchor inflation and inflation expectations. One interpretation of this unfavorable outcome is that monetary-fiscal cooperation can only be used effectively if the central bank has sufficient independence to credibly say that the period of higher inflation will be temporary, resulting in a higher price level, but not persistently higher inflation. However, the experience in Japan in recent decades suggests that an independent central bank, at least if it is believed to prefer low and stable inflation, may find it difficult to credibly commit to even a temporary period of higher inflation. Japanese prices fell fairly steadily for more than a decade starting in the late 1990s, and the economy expanded only slowly, on balance. A number of efforts to use fiscal and monetary policy proved insufficient to move the economy firmly out of deflation, and put growth on a firmer track. As a consequence, Japanese gross central government debt is now approximately 250 percent of GDP – very high under any interpretation – and inflation remains below target, despite highly accommodative monetary policy that has pushed BOJ assets to over 90 percent of GDP.26 In response to this difficult policy situation, the Japanese government has implemented a mix of more-accommodative monetary policy and highly accommodative fiscal policy.27 With regard to monetary policy, the BOJ has cut its short-term policy rate to negative territory, and it has indicated that it will conduct purchases of longer-term government securities in order to keep the yield on 10-year Japanese government bonds near zero.28 In addition, the Policy Board has indicated that it will maintain the low level of interest rates until inflation runs persistently above its 2 percent target rate.29 26 See FRED: https://fred.stlouisfed.org/series/GGGDTAJPA188N 27 In addition to structural policies – see Cabinet Office, Ministry of Finance, and Bank of Japan (2013). 28 The BOJ has also undertaken purchases of corporate instruments and JREITS andprovided low-cost financing for bank loans. See BOJ (2017). 29 See BOJ (2016). 28

This combination of policies has some attributes like those of the limited forms of monetaryfiscal cooperation we considered above. In particular, the overshoot of the inflation target could lead to a higher price level than would otherwise have been expected given the 2 percent inflation target,andtheimpliednegativerealinterestrateonlonger-termgovernmentdebtshouldreducethe pressure on fiscal policymakers and make possible a more accommodative fiscal policy than might otherwise be desirable. Indeed, some commentators have suggested that the highly accommodative monetary policy has allowed for more fiscal expansion, and one has even suggested that the policy mix has the “effect of helicopter money.”30 It will be interesting to see if these announced policies will be effective in finally bringing the long period of slow growth and very low inflation in Japan to an end. To summarize, the historical cases examined here suggest that to make monetary-fiscal cooperation effective the central bank needs to be able to credibly raise the expected price level, but without losing control of the price level. Control could be maintained by a credible commitment to the gold standard over the medium term, as in the United States during and after the Civil War. Even without a gold standard, the central bank may have sufficient credibility to contain inflation expectations, as in the United States after World War II. In addition, a gold standard or other similar commitment device may help to communicate an intended increase in the price level, making the higher price level credible while at the same time limiting expectations of a further rise in prices once the desired level is achieved. However, to keep price expectations anchored after the devaluation,itislikelyimportantthatthedevaluationbeseenbythepublicasaveryunusualevent that will not be repeated any time soon. That sort of expectation is probably easier to convey if the monetary-fiscal cooperation is undertaken in response to extraordinary circumstances, as was the case in France after World War I and Japan in the 1930s. Finally, without a framework for 30 Kihara (2016) quotes Etsuro Honda as saying that BOJ policy “in a sense, ...has the same effect of helicopter money.” Such an assessment seems overstated, however, since the BOJ is not explicitly monetizing government borrowing to finance fiscal policy. 29

communicating about the future level of prices, it may be very difficult to generate the desired outcome for one of two reasons. First, as in the United Kingdom in the 1960s, if the central bank is not independent and conditions are not extraordinary, then the change in the price level may not be seen as a one-off policy step, but rather as the start of a policy of permanently higher inflation. Alternatively, as in Japan in recent years, an independent central bank with a track record of low and stable inflation may find it difficult to credibly commit to even a temporary increase in inflation above the levels it is believed to prefer, potentially making the cooperative policy relatively ineffective. 5. Concluding Remarks The simulations in this paper suggest that a money-financed fiscal expansion, if understood and seen as credible by the public, could provide very substantial stimulus. A commitment to use monetary policy to boost nominal currency holdings by enough to finance the fiscal action would increase the effects of the combined action well beyond those of debt-financed fiscal stimulus alone. In particular, the change in monetary policy would increase the expected future price level, and thus reduce real interest rates substantially. However, as discussed above, policymakers may find the economic effects of such a program to be undesirable because of the large (and unpredictable) effects on inflation, and because it would risk unanchoring longer-term inflation expectations that could require costly actions to rectify. Moreover, making such a change in monetary policy clear and credible – outside of wartime and depressions – would likely be very difficult, and the macroeconomic effects could well prove to be verylimitedinpractice. Ifmonetarypolicyisconstrainedbytheeffectivelowerboundonthepolicy rate,thechangeinmonetarypolicycouldnotbesignaledbyachangeinthecurrentpolicyrate,but rather only by unconventional policy steps, such as asset purchases or providing forward guidance 30

regarding future short-term nominal rates. In addition, the announced change in monetary policy mightnotbetimeconsistent, andpolicymakerswouldlikelyfinditverydifficulttoconstrainfuture monetarypolicydecisions. Indeed,withcentralbankshavingspentconsiderabletimebuildingtheir reputations for low and stable inflation, the public would probably anticipate that policymakers would not follow through with the announced change in policy once the economy had recovered sufficiently. Moreover, by increasing uncertainty, the announcement of such a radical change in policycouldevenboostriskpremiumsandsobecounterproductive. Andifthechangeinmonetary policy only became credible gradually, as in our model with learning, then the impact of the new policy on the ecoomy would come later, after the economic recovery was well underway, at which point the higher inflation would not offer any benefit. That possibility could further undermine the credibility of the program, since it would likely heighten time-inconsistency problems. The credibility of the change in monetary policy might be strengthened if such a change were relatively modest – along the lines of the policies considered in section 3.3 – because it would not require a drastic change in the monetary policy framework, with all of the risks that such a change would entail. Such a limited approach would reduce the effects of the change on inflation and so make the policy less time inconsistent, while still supporting the effects of the fiscal action. Coordinated changes in monetary and fiscal policy could also help to make the change in monetary policy more credible than it would be on its own. Such a coordinated change could be seen as a signal that the authorities were committed to taking bold action to address a persistently weak economy and were willing to accept the consequences. In addition, expansionary fiscal policy, bystimulatingoutputandinflation,wouldbringforwardthedatewhenmonetarypolicywouldhave tightened in the absence of the associated change in monetary policy. Thus, the fiscal expansion couldhelpmakethechangeinmonetarypolicyvisibletothepublicsooner,enhancingthecredibility ofthechange. Moreover,governmentactiononthefiscalcomponentoftheprogramwouldallowthe 31

government to signal its agreement with the proposed change in monetary policy and acknowledge the higher inflation that it would imply. That signaling might also include an indication that the government anticipates that higher inflation will help keep the debt-to-income ratio in check, which would provide additional support for the credibility of the coordinated change in monetary policy. Of course, such steps on the part of the government could call into question the independence of the central bank. Indeed, the central bank could, in the future, be pressed to engage in such coordinated policies in situations when they were not warranted (that is, when output is not persistently low, inflation is not below target, and the policy interest rate is not constrained by its effective lower bound). However, even in the extreme case of a money-financed program, it may be possible to develop approaches that would help preserve monetary policy independence. For example, Bernanke (2016) has suggested a “dual key” approach in which the Federal Reserve would authorize the placement of funds in a special account that the Treasury could use to finance spending. However, the Congress would have to authorize the spending of funds in that account. Thus, both the Federal Reserve and the Congress would have to independently agree that a moneyfinanced fiscal program was desirable before it could be undertaken. A final issue, which we do not try to address here, is the possible international implications of money-financed fiscal programs. In particular, some recent research suggests that when monetary policyisconstrained bythe effectivelowerbound, moreaccommodative policy inoneeconomymay have adverse implications for other economies through its effects on exchange rates and trade.31 Thus, implementation of a money-financed fiscal program in one country could hurt its trading partners and potentially lead to the implementation of such programs in those economies as well. Such an outcome would not be undesirable in some circumstances, but given the possible spillovers, it might be appropriate to consult with policymakers in other jurisdictions before implementing 31 See,forexample,Caballero,Farhi,andGournchas(2016)andEggertson,Mehtotra,Singh,andSummers(2016) for a discussion of the international implications of monetary policy at the lower bound. 32

such a program. Doing so could help to avoid disorderly outcomes in foreign exchange markets. 33

References Bank of Japan (2013), “The ‘Price Stability Target’ under the Framework for the Conduct of Monetary Policy,” press release, January 22. (2016), “New Framework for Strengthening Monetary Easing: ‘Quantitative and Qualitative Monetary Easing with Yield Curve Control,”’ September 21. (2017). “Statement on Monetary Policy,” press release, January 31. Bernanke, Ben S. (2010). “The Economic Outlook and Monetary Policy,” speech at the Federal Reserve Bank of Kansas City Economic Symposium, Jackson Hole, Wyoming, August. (2016). “What tools does the Fed have left? Part 3: Helicopter money,” blog post, Brookings Institution, April 11. Blanchard, Olivier, and Roberto Perotti (2002). “An Empirical Characterization of the Dynamic Effects of Changes in Governement Spending and Taxes on Output,” Quarterly Journal of Economics, vol. 117(4), 1329-1368. Board of Governors of the Federal Reserve System (1940). Annual Report. Washington: Board of Governors of the Federal Reserve System (1946). Annual Report. Washington: Board of Governors of the Federal Reserve System Bordo, Michael D. and Pierre-Cyrille Hautcoeur (2003), “Why didn’t France follow the British Stabilization after World War One ?” mimeo., Rutgers University, May. Borio, Claudio, PitiDisyatat, AnnaZabai(2016). “Helicoptermoney: Theillusionofafreelunch,” blog post, Vox, May. 34

Buiter, Willem H. (2014). “The Simple Analytics of Helicopter Money: Why It Works – Always,” Economics, vol. 8, August. Caballero, Riccardo, Emmanuel Farhi, and Pierre-Olivier Gourinchas (2016). “Safe Asset Scarcity and Aggregate Demand,” mimeo., MIT, January. Cabinet Office, Ministry of Finance, and Bank of Japan (2013), “Joint Statement of the Government and the Bank of Japan on Overcoming Deflation and Achieving Sustainable Economic Growth,” January 22. Cairncross, Sir Alec (1996). Managing the British Economy in the 1960s: A Treasury Perspective. Basingstoke, U.K.: MacMillian Palgrave. Cecchetti, Stephen and Kermit Schoenholtz (2016). “A Primer on Helicopter Money,” blog post, June. Chaurushiya, Radha, and Ken Kuttner (2003), “Targeting the Yield Curve: The Experience of the Federal Reserve, 1942-51,” Memorandum to the Federal Open Market Committee, June 18. Christiano, Lawrence, Martin Eichenbaum, and Sergio Rebelo (2011). “When is the Government Spending Multiplier Large?” Journal of Political Economy, vol. 119(1), 78-121. Eggertson, Gauti B. and Michael Woodford (2003). “The Zero Bound on Interest Rates and Optimal Monetary Policy,” Brookings Papers on Economic Activity, no. 1, 139-211. , Neil R. Mehrotra, Sanjay R. Singh, and Lawrence H. Summers (2016). “A Contagious Malady? Open Economy Dimensions of Secular Stagnation,” NBER Working Paper No. 22299. June. Eichengreen, Barry and Peter M. Garber (1991). “Before the Accord: U.S. Monetary-Financial 35

Policy, 1945-51,” chapter 5 in Hubbard, R. Glenn, ed., Financial Markets and Financial Crises, Chicago: University of Chicago Press for the NBER. Engen, Eric M., Thomas Laubach, and David Reifschneider (2017). “The Macroeconomic Effects of the Federal Reserve’s Unconventional Monetary Policies,” Federal Reserve Board, Finance and Economics Discussion Series, No. 2015-005. English, William B., David Lopez-Salido, and Robert J. Tetlow (2013). “The Federal Reserve’s Framework for Monetary Policy: Recent changes and new questions,” presented at the 14th Jaques Polak Annual Research Conference, November. Erceg, ChristopherJ.andJesperLinde(2014). “IsThereaFiscalFreeLunchinaLiquidityTrap?” Journal of the European Economic Association, vol. 12 (1), 73-107. Farhi, Emmanuel and Ivan Werning (2012) “Fiscal Multipliers: Liquidity Traps and Currency Unions,” mimeo. , Harvard University. Fischer, Stanley (1982). “Seigniorage and the Case for a National Money,” Journal of Political Economy, vol. 90(2), 295-313. Friedman, Milton (1969). The Optimum Quantity of Money and Other Essays, Chicago: Aldine Publishing Company. Gali, Jordi (2016). “The Effects of a Money-Financed Fiscal Stimulus,” mimeo., CREI, December. Ireland, Peter N. (2014). “The Macroeconomic Effects of Interest on Reserves,” Economic Dynamics, vol. 18(6), 1271-1312. Kihara, Leika (2016), “Japan sinking deeper into de-facto helicopter money,” Reuters, August 19. (http://www.reuters.com/article/us-japan-economy-boj-helicopter-analysisidUSKCN10U0DS?il=0) 36

Neslon, Edward and Kalin Nikolov (2003). “UK Inflation in the 1970s and 1980s: The role of output gap mismeasurement,” Journal of Economics and Business, vol. 55(4), 353-370. Ramey, Valerie A. and Sarah Zubairy (forthcoming). “Government Spending Multipliers in Good Times and in Bad: Evidence from U.S. Historical Data,” Journal of Political Economy. Reichlin, Lucrezia, Adair Turner, and Michael Woodford (2013). “Helicopter money as a policy option,” VoxEu, 20 May. Rockoff, Hugh (2014), “War, Money, and Inflation in the United States from the Revolution to the Vietnam War,” mimo., Rutgers University. Romer, Christina (2014). “It Takes a Regime Shift: Recent Developments in Japanese Monetary Policy through the Lens of the Great Depression,” NBER Macroeconomics Annual 2013, vol. 28, pp. 383-400. Shizume, Masato (2009). “The Japanese Economy during the Interwar Period: Instability in the Financial System and the Impact of the World Depression,” Bank of Japan Review, May, pp. 1-10. Sutch, Richard and Susan B. Carter, Eds. Historical Statistics of the United States, Millennial Edition Online, Cambridge: Cambridge University Press. Taylor, John (1993). “Discretion Versus Policy Rules in Practice,” Carnegie-Rochester Conference Series on Public Policy, vol. 39, 195-214. Turner, Adair (2015). “The Case for Monetary Finance: An essentially political issue,” presented at the 16th Jaques Polak Annual Research Conference, November. Woodford, Michael (2003). Interest and Prices: Foundations of a Theory of Monetary Policy, Princeton, NJ: Princeton University Press. 37

Woodford,Michael(2012). “MethodsofPolicyAccommodationattheInterest-RateLowerBound,” paper presented at the Federal Reserve Bank of Kansas City Symposium on “The Changing Policy Landscape,” Jackson Hole, Wyoming, August 3 (Revised September 16). 38