Modeling Fiscal Matters

Transcription

1 Modeling Fiscal Matters Eric M. Leeper Department of Economics, Indiana University October 21 Kansas City Fed Forecasting Workshop

2 The Messages Key aspects in modeling fiscal policy: 1. expectations 2. long-lasting dynamics 3. information (fiscal foresight) 4. interactions with monetary policy 5. nonlinearity 6. uncertainty

3 Recent Macro Policies Monetary and fiscal policy responses to recession and financial crisis of have been unusual aggressive United States, Japan, China, many European countries employed large discretionary fiscal stimulus packages Many central banks have driven interest rates to near zero and engaged in unconventional operations that have exploded their balance sheets This lecture pulls together those key features of fiscal policy to address potential consequences of these actions Draws on Leeper-Plante-Traum (21), Leeper-Walker-Yang (21), Davig-Leeper (21), Bi (29), Bi-Leeper (21)

4 The Messages Estimates of fiscal stimulus depend strongly on how stimulus is implemented tax cuts (which taxes); spending increases (which spending) how and when the private sector expects the resulting debt expansion will be financed whether the stimulus occurs gradually, so agents have fiscal foresight how monetary policy behaves whether it is active or passive Unfortunately, many of these considerations play little role in government projections of impacts of fiscal stimulus

5 The U.S. Example American Reinvestment and Recovery Act: $787 Billion (5 % GDP) Financed with new government debt issuance Rationale provided by paper by Romer-Bernstein reporting multipliers for permanent 1% of GDP increase in G and decrease in T forecasts of unemployment rate with and with stimulus claim GDP will be 3.7% higher; 3.6 million new jobs

6 Romer-Bernstein Multipliers GDP Multiplier for Higher Government Purchases Quarter 1.8 GDP Multiplier for Lower Taxes Quarter Permanent Fiscal Shocks

7 Some Questions What economic models underlie the multipliers? Are the numbers reproducible? Why consider permanent changes when the Act makes transitory changes? What are the consequences of the stimulus for government debt? What are the repercussions of significantly higher debt? Will the debt run-up be sustained or retired? At what level will debt stabilize? How will policies adjust in the future to either sustain or retire debt? What assumptions about current and future monetary policy are embedded in the multipliers?

8 Some Answers from Obama Administration

9 Some Answers from Economic Research Four models of fiscal policy 1. Neoclassical growth model I (Leeper-Plante-Traum) fiscal detail: 3 taxes rates, G consumption, transfers estimated to U.S. data 2. Neoclassical growth model II (Leeper-Walker-Yang) fiscal detail: 2 tax rates, G consumption, G investment, transfers time-to-build in government infrastructure foresight estimated to U.S. data 3. New Keynesian model (Davig-Leeper) monetary & fiscal policy with regime switching in policies calibrated to U.S. data 4. Model of sovereign debt default (Bi) stochastic Laffer curve & fiscal limit nonlinear risk premia

10 Some Answers from Economic Research There is also a ton of VAR evidence on multipliers Variety of identification schemes restrictions on elasticities and timing (Blanchard-Perotti) restrictions on signs of impulse responses (Mountford-Uhlig) Caldara & Kamps show fiscal VARs are generically unidentified: ultimately, identification achieved by ad hoc additional restrictions Joonyoung Kim is finding that two fresh kinds of restrictions have bite 1. intertemporal government budget constraint 2. combined with sources of fiscal financing The presumed death of VARs may be premature

11 Neoclassical Growth Model I Conventional except for specification of policy behavior tax rules ˆτ t k = ϕ k Ŷ t + ˆB γk t 1 + φ kl u l t + φ kc u c t + u k t ˆτ t l = ϕ l Ŷ t + γ lˆb t 1 + φ lk u k t + φ lc u c t + u l t ˆτ t c = φ kc u k t + φ lc u l t + u c t spending rules Ĝ t = ϕ g Ŷ t γ gˆb t 1 + u g t Ẑ t = ϕ Z Ŷ t γ Z ˆB t 1 + u z t hats are log-deviations, u s are AR(1) with innovations N(, 1)

12 Growth Model I: Results Data like to have many instruments adjust to stabilize debt Multipliers tend not to be very large caveat: with certain monetary policies, multipliers can be much larger Short-run and long-run multipliers can be very different Source of financing can matter a lot, especially at longer horizons Both speed at which debt stabilized and size of automatic stabilizers ϕ s matter for fiscal impacts Takes many years to establish present-value budget balance 2 or more

13 Fiscal Multipliers A common measure [Blanchard-Perotti (22), Romer-Bernstein (29)] Impact Multiplier(k) = Y t+k G t Sweeps dynamics of fiscal variables under the rug Present value multiplier [Mountford and Uhlig] Present Value Multiplier(k) = k E t j= i= k E t j= i= j (1 + r t+i ) j Y t+k j (1 + r t+i ) j G t+k

14 Growth Model I: Multipliers Capital Tax Present-Value Multipliers Variable 1 quarter 1 quarters PV( Y) PV( T k ) PV( C) PV( T k ) Labor Tax Present-Value Multipliers Variable 1 quarter 1 quarters PV( Y) PV( T l ) PV( C) PV( T l ) All fiscal instruments respond to debt

15 Growth Model I: Multipliers Capital Tax Present-Value Multipliers Variable 1 quarter 1 quarters PV( Y) PV( T k ) PV( C) PV( T k ) Labor Tax Present-Value Multipliers Variable 1 quarter 1 quarters PV( Y) PV( T l ) PV( C) PV( T l ) Only capital and labor taxes respond to debt (red)

16 Growth Model I: Multipliers Government Spending Present-Value Multipliers Variable 1 quarter 1 quarters PV( Y) PV( G) PV( C) PV( G) Transfers Present-Value Multipliers Variable 1 quarter 1 quarters PV( Y) PV( Z) PV( C) PV( Z) All fiscal instruments respond to debt

17 Growth Model I: Multipliers Government Spending Present-Value Multipliers Variable 1 quarter 1 quarters PV( Y) PV( G) PV( C) PV( G) Transfers Present-Value Multipliers Variable 1 quarter 1 quarters PV( Y) PV( Z) PV( C) PV( Z) Only capital and labor taxes respond to debt (red)

18 Government Spending Multipliers Output Multipliers.6.4 All instruments adjust Quarters After an Increase in Government Consumption

19 Government Spending Multipliers Output Multipliers.6.4 All instruments adjust $1 more government spending $.65 more GDP Quarters After an Increase in Government Consumption

20 Government Spending Multipliers Output Multipliers.6.4 Only transfers adjust Quarters After an Increase in Government Consumption

21 Government Spending Multipliers Output Multipliers.6.4 Only transfers adjust If higher spending financed with lower transfers, GDP rises more Quarters After an Increase in Government Consumption

22 Government Spending Multipliers Output Multipliers Government spending adjusts Quarters After an Increase in Government Consumption

23 Government Spending Multipliers Output Multipliers Government spending adjusts.2.4 If government spending financed by lower government spending, GDP falls after 2 years Quarters After an Increase in Government Consumption

24 Government Spending Multipliers.8 Output Multipliers.6.4 Taxes adjust Quarters After an Increase in Government Consumption

25 Government Spending Multipliers.8 Output Multipliers.6.4 Taxes adjust If government spending financed by higher taxes, GDP soon begins to decline Quarters After an Increase in Government Consumption

26 Speed of Fiscal Adjustment Obama administration has pledged to cut deficit in half within 4 years Echoing Europe, where cuts are actually occurring Done in response to outcries about fiscal unsustainability Use estimated model to answer: What are the implications for effectiveness of fiscal stimulus of slowing down or speeding up fiscal adjustments? slowing down pushes adjustments into future rational agents discount those more heavily speeding up brings them forward Changes in the timing of fiscal adjustments can alter the government spending multipliers in important ways

27 Speed of Adjustment of Fiscal Instruments Modify fiscal rules to vary responsiveness to debt tax rules ˆτ t k = ϕ k Ŷ t + ˆB µγk t 1 + φ kl u l t + φ kc u c t + u k t ˆτ t l = ϕ l Ŷ t + µγ lˆb t 1 + φ lk u k t + φ lc u c t + u l t ˆτ t c = φ kc u k t + φ lc u l t + u c t spending rules Ĝ t = ϕ g Ŷ t µγ gˆb t 1 + u g t Ẑ t = ϕ Z Ŷ t µγ Z ˆB t 1 + u z t vary µ to speed up or slow down adjustment

28 Government Spending Multipliers.7 Output Multipliers Historically Estimated Speed of Adjustment Quarters After an Increase in Government Consumption

29 Government Spending Multipliers.7 Output Multipliers Slower Speed of Adjustment Quarters After an Increase in Government Consumption

30 Government Spending Multipliers Output Multipliers Slower retirement of debt enhances fiscal stimulus for 6 years.3 Slower Speed of Adjustment Quarters After an Increase in Government Consumption

31 Government Spending Multipliers.7 Output Multipliers Faster Speed of Adjustment Quarters After an Increase in Government Consumption

32 Government Spending Multipliers Output Multipliers Faster retirement of debt suppresses fiscal stimulus.3 Faster Speed of Adjustment Quarters After an Increase in Government Consumption

33 Strength of Automatic Stabilizers.5.5 No automatic stabilizers Present-value G multipliers for output: varying ϕ s

34 Strength of Automatic Stabilizers.5 No automatic stabilizers.5 Estimated automatic stabilizers Present-value G multipliers for output: varying ϕ s

35 Strength of Automatic Stabilizers.5 No automatic stabilizers.5 1 Estimated automatic stabilizers 2 estimated automatic stabilizers Present-value G multipliers for output: varying ϕ s

36 Strength of Automatic Stabilizers.5 No automatic stabilizers.5 1 Estimated automatic stabilizers 2 estimated automatic stabilizers 3 estimated automatic stabilizers Present-value G multipliers for output: varying ϕ s

37 Neoclassical Growth Model II In U.S. and Europe, heavy emphasis on government infrastructure spending Similar in structure to previous model; two important extensions introduction of productive government investment G I introduction of time-to-build in government capital Distinguish between budget authority and outlays authority occurs first, giving total spending and planned path of outlays implementation delays modeled with time-to-build

38 Implementation Delays: Example Estimated costs for highway construction in Title XII of the American Recovery and Reinvestment Act of Tota Budget Authority Estimated Outlay Billions of dollars. Source: Congressional Budget Office

39 Modeling Government Investment Aggregate production Y t = A (u t K t 1 ) α K (L t ) α L ( K G t 1 ) αg αg critical (α G = unproductive) A I t: budget authorization; N quarters to complete project Law of motion for public capital K G t = (1 δ G ) K G t 1 + A I t N+1 budget authorization process an AR(1) Government investment implemented at t (outlaid) N 1 G I t = φ n A I t n, n= N 1 φ = 1; φ s are outlay rates

40 Role of Government Productivity C I L Y α G =.1 α G =.5 α G = years I L C Y years Permanent shock Temporary shock No implementation delays and lump-sum financing

41 Implementation Delays and Foresight C C I 1 quarter 1 year 3 years L Y G I.1 1 quarter 3 years.5 1 year I L Y G I years years α G =.1 α G =.5 With implementation delays

42 New Keynesian Model Two key distortions that given monetary policy real effects: monopolistic competition sluggish price adjustment Elastic labor supply; inelastic capital Transmission mechanism of MP: real interest rates Transmission mechanism of FP: real interest rates & wealth effects Integrate monetary and fiscal policy interest rate rule for MP exogenous process for government spending lump-sum taxes

43 New Keynesian Model Estimate switching rules for monetary & tax policy Embed rules in calibrated model Four possible policy regimes: 1. Active MP/Passive FP 2. Passive MP/Active FP 3. Passive MP/Passive FP 4. Active MP/Active FP With fixed regime: Passive/Passive indeterminacy With fixed regime: Active/Active non-existence Can study consequences of periodically visiting those forbidden regimes Focus on effects of unproductive G

44 U.S. Policy Responses to Recession Unusually aggressive joint policy response federal funds rate near zero bound since Dec 8 Fed s balance sheet has more than doubled: $8 billion to $2.5 trillion $125 billion tax refund in 8 and $787 billion stimulus package in 9 deficit is 13% of GDP now; debt will rise from 4% to 8% of GDP over the decade; may reach 277% by 24 Objective of stimulus is to create jobs by increasing consumption demand, labor demand, employment

45 The Modeling Effort Model two aspects of the policy response 1. joint monetary and fiscal policy effort 2. current aggressive policies not likely to continue indefinitely Use standard new Keynesian model with monetary and fiscal policy regime change Bottom-line: government spending multipliers can be large or small, depending on policy regime Simulate effects of American Recovery and Reinvestment Act under alternative policy assumptions

46 Government Spending: Crowd Out or In? Policy Research Romer-Bernstein: output multiplier 1.5 and very persistent CBO: stimulus makes recession less severe and shorter lived no professional consensus that higher G raises private C RBC or standard new Keynesian models G crowds out C empirical evidence mixed, but favors crowding in

47 Policy Regimes Since the late 194s, U.S. monetary & fiscal policies have fluctuated among: Active MP Taylor principle holds Passive MP Taylor principle not satisfied Passive FP PV of taxes = PV of G Active FP PV of taxes < PV of G Current policy: passive MP & active FP

48 Why Policy Regime Matters Following an increase in G Passive MP allows the real interest rate to fall in response to higher expected inflation 2. Active FP diminishes the negative wealth effect induced by higher taxes Both of these increase the stimulative effect of government spending These do not happen under the usual active MP/passive FP regime A natural & relevant way to get large G multipliers

49 Monetary Policy Rule Estimates The monetary policy rule is r t = α (S M t ) + α π (S M t )π t + α y (S M t )y t + σ r (S M t )ε r t S M t follows a four-state Markov chain reaction coefficients and shock volatility switch independently Monetary policy breaks into regimes with A strong response to inflation (active): απ = 1.29 A weak response to inflation (passive): απ =.53

50 Fiscal Policy Rule Estimates The fiscal policy rule is τ t = γ (S F t ) + γ b (S F t )b t 1 + γ y (S F t )y t + γ g (S F t )G t + σ τ (S F t )ε τ t S F t follows a two-state Markov chain Fiscal policy breaks into regimes with Taxes rise in response to debt (passive): γb =.7 Taxes fall in response to debt (active): γb =.25

51 U.S. Monetary and Fiscal Regimes AM,PF Ricardian AM,AF Explosive PM,PF Indeterminacy PM,AF Fiscal Theory

52 Model Setup We use a basic New Keynesian model with variable government purchases fixed capital; elastic labor supply; Calvo price rigidities Unproductive government spending financed via: lump-sum taxes; one-period nominal bonds; seigniorage revenues Government purchases follow AR(1) (for now...) Government demands goods in same proportion as private sector

53 Perspective on Transmission of G The ubiquitous Intertemporal Equilibrium Condition holds in all regimes M t 1 + (1 + r t 1 )B t 1 P t = E t T=t A government liabilities valuation equation ( [q t,t τ T G T + r )] T M T 1 + r T P T Higher path for G without an equivalent higher path for τ lowers the present value of primary surpluses creates an imbalance at initial prices between the value of debt and its expected backing Equilibrium restored via a higher path of P, which is consistent with firms raising prices

54 Higher G: Active MP / Passive FP % basis points basis points Output Gap Inflation Nominal R 2 % Gov Purchases % basis points Consumption.6 AM/PF Real Rate 5 % % Debt (level) Taxes

55 Higher G: Passive MP / Active FP % basis points basis points Output Gap Inflation Nominal R 2 % Gov Purchases % basis points Consumption.6 AM/PF.4 PM/AF Real Rate 5 % % Debt (level) Taxes

56 Intertemporal Adjustments 2 Debt (level) 1 Primary Surplus PV Primary Surplus PV Seigniorage AM/PF

57 Intertemporal Adjustments 2 Debt (level) 1 Primary Surplus PV Primary Surplus PV Seigniorage AM/PF PM/PF

58 Intertemporal Adjustments 2 Debt (level) 1 Primary Surplus PV Primary Surplus PV Seigniorage AM/PF PM/PF PM/AF

59 Present Value Multipliers PV( Y) PV( G) after Regime 5 quarters 1 quarters 25 quarters AM/PF PM/PF PM/AF Table: Note: PV( C) PV( G) = PV( Y) PV( G) 1 Values greater than unity imply a positive consumption response to increases in G

60 Simulating Stimulus: The 29 ARRA The 29 ARRA includes around $35 billion in spending on infrastructure, energy, healthcare, etc. $144 billion in federal transfers to state and local governments Following Romer and Bernstein assume 6 percent is devoted to new spending We use the same path for additional G as Cogan, Cwik, Taylor, Wieland Simulate under different monetary-fiscal combinations

61 The ARRA s Path for G The Fiscal Stimulus: Path of Government Spending

62 29 ARRA: AM/PF % basis points Output Gap Inflation 3 % % Gov Purchases Debt 4 2 AM/PF % basis points Consumption Real Rate 5 % % Taxes Primary Surplus

63 29 ARRA: AM/PF & PM/AF % basis points Output Gap Inflation 3 % Gov Purchases 4 2 PM/AF AM/PF Debt 4 % basis points Consumption Real Rate 5 % Taxes Primary Surplus 2 % %

64 A Risky Game of Chicken What if, as inflation begins to rise, the Fed switches to an active stance (from PM/AF)? This is a very real possibility when there is no coordination between MP & FP Then there are two unstable relationships: inflation due to the active MP debt due to the active FP In a fixed AM/AF regime, there would be no equilibrium With switching, so long as you are sufficiently far from the fiscal limit, there is a build up of debt And persistently higher inflation because MP has lost control of inflation

65 The 29 ARRA: Active/Active % basis points Output Gap AM/PF Inflation 3 % % Gov Purchases Debt 1 5 PM/AF AM/AF % basis points.8.4 Consumption Real Rate 5 % % Taxes Primary Surplus

66 Nonlinearity & Fiscal Policy Fiscal limits are country specific: depend on government size, degree of countercyclical fiscal policy, political risk, and shock processes Risk premia are nonlinear in level of government debt Long-term bonds can provide early warning Fiscal reforms can significantly shift distribution of fiscal limits

67 Recent Sovereign Risk Premia Ireland Greece Spain Italy Portugal Long term Interest Rate Spread over Germany Apr 8 Jun Aug Oct Dec Feb 9 Apr Jun Aug Oct Dec Feb 1 Apr

68 A Model Exogenous technology and government spending: ln At A ln gt g Household problem: s.t. FOC: = ρ u ln A t 1 A + εa t ε A t N (, σ 2 A ) = ρ e ln g t 1 g + ε g t ε g t N (, σ 2 g ) max E β t u (c t, L t ) t= A t (1 τ t )(1 L t ) + z t c t = b t q t (1 t )b t 1 }{{} b d t u L (t) = A t (1 τ t ) u c (t) [ q t = βe t (1 t+1 ) u ] c(t + 1) u c (t)

69 A Model Government budget: τ t A t (1 L t ) + b t q t = g t + z t + (1 t )b t 1 }{{} b d t Unenforceable bond contract: { if bt 1 < b t = t with b t N (b, σb 2) δ if b t 1 b t Debt-stabilizing tax rule: τ t τ = γ ( ) b d t b Countercyclical lump-sum transfers: ln z t z = ζz ln A t A

70 Dynamic Laffer Curve T t = τ t A t (1 L t ) => T max (A, g) = T (τ max (A, g); A, g) Tax Revenue Tax Rate Productivity 1.15

71 Fiscal Limit Fiscal limit: maximum sustainable level of government debt B u max c (t) = E u max c () }{{} θ t }{{} (T max t g t z t ) }{{} t= political risk future max fiscal surplus discount rate The distribution depends on: Government size: g/y and z/y Countercyclical lump-sum transfers: ζ z Political risk: < θ t 1 (ICRG index) Standard & Poor s (28): stability, predictability, and transparency of a country s political institutions are important considerations... Shock processes MCMC simulation: Simulate N paths to approximate N (b, σ 2 b ).

72 Fiscal limit: Simulation 1 5 Government Purchases GDP g/y=.29 g/y=.213 g/y= Lump sum Transfers GDP z/y=.224 z/y=.157 z/y= Debt GDP Countercyclicality 15 ζ z = ζ z = ζ z = Debt GDP Shock Persistence of A 15 ρ A = ρ A = ρ A = Debt GDP Shock Persistence of g 1 ρ g = ρ g =.553 ρ g = Debt GDP Debt GDP Political Risk 15 θ=.96 1 θ=.83 5 θ= Debt GDP Shock Standard Deviation of A 15 σ A =.34 1 σ A =.2 5 σ A = Debt GDP Shock Standard Deviation of g 1 σ g = σ g =.2 σ g = Debt GDP

73 Fiscal limit: Data AAA New Zealand 12 AAA Canada 12 Rating AA+ AA Debt GDP AA AA Italy 14 AA Belgium 14 Rating AA AA AAA Debt GDP A+ AAA Sweden 2 AAA Japan 2 Rating AA+ AA AA Debt GDP AA Year Year

74 Nonlinear solution Monotone mapping method (Coleman (1991), Davig (24)): q t = βe t ( (1 t+1 ) uc(t + 1) ) u c(t) ) b d t + g t + z(ψ t) τ(ψ t)a t (1 L(ψ t) f b (ψ t) = βe t { ( 1 (f b (ψ t), b t+1 ) ) uc(f b (ψ t), A t+1, g t+1, b t+1 ) u c(ψ t) } (1) (2) Grid points of 3-dimension state space, ψ t = (b d t, g t, A t ), using Tauchen (1991) Initial guess of the decision rule f b (.) (b t = f b (ψ t)) Update the decision rule fi b (.) by iterating over equation (2) until it converges (ɛ = 1e 8) Numerical integration: Newton-Cotes formulas.

75 Calibration Default scheme: A higher uncertainty of fiscal limits implies higher δ t = { if bt 1 < b t δ 2σ b b if b t 1 b t (b t N (b, σ 2 b )) Calibrate to Greece ( ): τ L γ z/y ζ z g/y ρ g σ g θ H θ L p β L ρ A σ A / Markov switching θt : θ t {θ H, θ L } with p LL = p HH = p

76 Fiscal Limit: Greece High θ Markov θ Low θ Debt GDP

77 Decision Rule: R(b d, A, g) Net Interest Rate (r) A at ss Low A High A r(b) with g at Steady State under Different A b * 2std b * Bond (b) Net Interest Rate (r) g at ss Low g High g r(b) with A at Steady State under Different g b * 2std b * Bond (b)

78 Simulation: A Severe Recession Given the paths of A t and g t. At each period, the effective fiscal limit (b t, green line) is drawn from the approximated distribution. The paths of c t, L t, τ t, b t, r t are determined by equilibrium conditions. t=1 t=2 t=3 t=4 t=5 t= 6 A t -4.88% -8.61% -9.97% -6.67% -4.21% -1.92% g t /y t 2.35% 21.68% 21.81% 21.8% 2.29% 19.52%

79 Nonlinear Simulation.26 Labor (1 L t ) Tax Rate (τ t ) Bond (b t 1 ) Consumption (c t ) Transfer (S t ) Net Interest Rate (r t 1 ) Default free Default Government Spending (g t ).48 1 Productivity (A t ).6 Default Probability

80 Long-term Bonds Price of long-term bond with maturity n: ( Q n t = β n E t (1 t+n ) u ) c(t + n) u c (t) r n t = 1 Q n t 1 Q nf t Solution: finite-element method

81 Simulation: Long-Term Bonds year bond 3 year bond 5 year bond 7 year bond 1 year bond

82 Wrap Up Modeling fiscal matters calls for substantial extensions to and modifications of existing DSGE models 1. long-run issues: linearizing around steady state? 2. nonstationarity: linearizing around steady state? 3. nonlinearity: linearizing around steady state? 4. nonnormality: linearizing around steady state? May be the death of Dynare