Fiscal Policy in an Estimated DSGE Model of the Japanese Economy

Transcription

1 Fiscal Policy in an Estimated DSGE Model of the Japanese Economy Do Non-Ricardian Households Explain All? Yasuharu Iwata Economic and Social Research Institute, Cabinet O ce, Government of Japan June 2009 Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 1 / 31

2 Motivation Studies on Bayesian estimation of DSGE models of the Japanese economy are growing (cf. Iiboshi, Nishiyama, and Watanabe (2008) (INW) and Sugo and Ueda (2008) (SU)), but almost no attention has been paid to scal policy. Inclusion of non-ricardian Households is currently the most popular way to generate a positive response of consumption. Seemingly contradictive observation in Japan: Hatano (2004): Japan s non-ricardian share stays in throughout the 1980s and the 1990s. Mahfouz, Hemming, and Kell (2002): Fiscal multipliers are larger for Japan than in the U.S. and Europe. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 2 / 31

3 This Paper Introduce non-ricardian households and three distortionary taxes to the Smets and Wouters (2003) model. Introduce feedback rules for each tax following Forni, Monteforte, and Sessa (2009) (FMS). Estimate the model via MCMC using Japanese scal data: The aggregate e ective tax rates are calculated following Mendoza, Razin, and Tesar (1994). Only quarterly data are utilized. Capital income tax series was, however, too volatile and not used. Examine the role of tax rules in scal policy e ectiveness by conducting scal policy simulations under various tax rule combinations. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 3 / 31

4 Main Findings Non-Ricardian share in Japan is smaller than those in the euro area and the United States. The model successfully delivers the crowding-in e ect on consumption regardless of its low share of non-ricardian households. Fiscal policy becomes more e ective if its nance is allocated lightly on labor-dampening taxes. A choice of tax rule combination can dominate the non-ricardian share in its e ect on scal multipliers. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 4 / 31

5 Model Features (Households (1)) Ricardian household i maximizes its lifetime utility: E t β t ε b 1 t 1 σ c Ct R (i) hc R 1 σc ε l t t 1 1+σ l L R t (i) 1+σ l t=0 subject to (1 + τ c t )Ct R (i) + I t (i) + Ψ(z t (i))k t 1 (i) + B t (i) R t P t = (1 τ d t )w t (i)l R t (i) + (1 τ k t )rt k z t (i)k t 1 (i) + (1 τ k t ) D t (i) K t (i) = (1 δ)k t 1 (i) + h 1 S ε i t I t (i) I t 1 (i) i I t (i). P t + B t 1(i) P t, Non-Ricardian household j simply consumes its after tax income: (1 + τ c t )Ct NR (j) = (1 τ d t )w t (j)l NR (j). Aggregate consumption: C t = (1 ω)ct R (i) + ωct NR (j). t Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 5 / 31

6 Model Features (Households (2)) Non-Ricardian households are assumed to set their wages equal to the average wage of Ricardian households. Both wages and labor hours will be equal for all households: Wt R (i) = Wt NR (j) = W t (n), L R t (i) = L NR t (j) = L t (n). Ricardian households reset their wages optimally with probability 1 ξ w, otherwise adjust them according to an indexation scheme Pt 1 Wt R γw (i) = P t 2 Wt R 1 (i). Aggregate nominal wage law of motion is then expressed as: W t = 2 4(1 ξ w ) (W t (n)) 3 1 γw λw 1,t λw,t + ξ Pt 1 w P t 2 W t 1 (n) 5 λ w,t. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 6 / 31

7 Model Features (Households (3)) W t (n) = Wt R (i) is chosen such that maximizes: E t s=0(βξ w ) s h 1 1 σ c C R t+s (i) hc R t+s 1 1 σc ε l t 1+σ l L t+s (i) 1+σ l subject to (1 + τ c t+s )C R t+s (i) + I t+s (i) + Ψ(z t+s (i))k t+s 1 (i) + B t+s (i) i R t+s P t+s = (1 τ d t+s ) W t R (i) P t+s L t+s (i) + (1 τ k t+s )rt+sz k t+s (i)k t+s 1 (i) + (1 τ k t+s ) D t+s (i) P t+s + B t+s 1(i) P t+s, where L t+s (i) = W R t (i) W t+s 1+λw,t+s λ w,t+s Lt+s. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 7 / 31

8 Model Features (Fiscal and Monetary Authorities) Government budget constraint: G t + B t 1 P t = τ c t C t + τ d t w t L t + τ k t r k t z t K t 1 + τ k t D t P t Fiscal policy feedback rules (log-linearized): ˆτ c t = ρ tc ˆτ c t 1 + (1 ρ tc ) φ tcb (ˆb t 1 Ŷ t 1 ) + η tc t, ˆτ d t = ρ td ˆτ d t 1 + (1 ρ td ) φ tdb (ˆb t 1 Ŷ t 1 ) + η td t, ˆτ k t = ρ tk ˆτ k t 1 + (1 ρ tk ) φ tkb (ˆb t 1 Ŷ t 1 ) + η tk t, Ĝ t = ρ g Ĝ t ρ g φ gy Ŷ t 1 + η g t. + 1 R t B t P t. Monetary policy feedback rule (log-linearized): ˆR t = ρ r ˆR t 1 + (1 ρ r ) φ r π ˆπ t 1 + (1 ρ r ) φ ry Ŷ t + η R t. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 8 / 31

9 Estimation Results Summary (Structural parameter means) Overall, the values of posterior mean estimates are not so di erent from those reported in previous studies. The estimated mean value of non-ricardian share (ω) 0.25 is very much consistent with the Kalman lter estimates of Hatano (2004). The Calvo parameters (ξ w = 0.82, ξ p = 0.43) are in line with the results of Koga and Nishizaki (2005). Parameter values for habit persistency (h = 0.47) and labor supply elasticity (σ l = 2.11) are in between INW and SU. All in all, posterior means of structural parameters do not suggest large scal multipliers in this economy. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 9 / 31

10 Estimation Results Summary (Policy parameter means) The estimated response of monetary policy to in ation is weak (ρ r = 0.93, φ r π = 1.53) but close to the results of Ichiue, Kurozumi, and Sunakawa (2008) (ρ r = 0.85, φ r π = 1.49). Posterior mean estimates of tax rule parameters (φ tcb = 0.013, φ tdb = 0.005, φ tkb = 0.123) suggest that capital income taxation played a central role in stabilizing the government debt in Japan. The estimated persistence parameters (ρ tc = 0.51, ρ td = 0.57, ρ tk = 0.66) are all substantially smaller than those of FMS (ρ tc = 0.96, ρ td = 0.91, ρ tk = 0.97). Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 10 / 31

11 Impulse Responses The model successfully delivers the crowding-in e ect on consumption, regardless of its relatively small non-ricardian share, price stickiness, and habit persistency Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 11 / 31

12 Fiscal Multipliers Japan Euro Area Quarters This paper ESRI FMS NAWM QUEST III by (1st yr) 0.28 (0.72) (1.00) (2nd yr) (-0.04) (0.06) n.a. bc n.a. n.a. 4 (1st yr) (0.05) (0.09) n.a. n.a. 8 (2nd yr) (-0.12) (0.23) n.a. n.a. bi n.a. n.a. 4 (1st yr) (0.07) (-0.73) n.a. n.a. 8 (2nd yr) (-1.06) (-0.24) n.a. n.a. Values in parentheses are yearly average e ects. Yearly e ects of a one year-long increase in government investment. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 12 / 31

13 Fiscal Multipliers Summary The output multiplier in the rst period is larger than those of other DSGE models for the euro area, re ecting the strong increase in investment and successfully crowded-in consumption. The yearly averages of the multipliers are not so di erent from those of Japan s ESRI model, regardless of DSGE model structure. In later periods, the model exhibits large decreases in consumption, investment, and, hence, in output. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 13 / 31

14 Policy Experiments (1) Under the estimated tax rules (baseline speci cation, φ tcb = 0.013, φ tdb = 0.005, φ tkb = 0.123), both consumption and investment multipliers in initial periods are larger than those under FMS speci cation (φ tcb = 0.041, φ tdb = 0.058, φ tkb = 0.050). It seems that investment serves as a major driving force for the stronger output response of baseline speci cation C (Baseline) I(Baseline) C (FMS spec.) I(FMS spec.) Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 14 / 31

15 Policy Experiments (2) Ricardian consumption shows a similar pattern to investment. The initial decline of Ricardian consumption in more anti-in ationary monetary policy case (φ r π = 1.7) is smaller than that of FMS speci cation: The rather greater scal multipliers of baseline owe much to its tax rules Baseline Specification 1 FM S spec. Anti inflationarypolicy Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 15 / 31

16 General Remarks The size of a coe cient on a debt-to-output ratio of a tax rule determines the timing of the taxation and accordingly a ects the time paths of economic variables. The initial decline of Ricardian consumption after a government spending shock in speci cation 1 (φ tcb = φ tdb = φ tkb = 0.2) is smaller than that in FMS speci cation (φ tcb = 0.041, φ tdb = 0.058, φ tkb = 0.050). The di erence between scal multipliers under baseline and those under FMS speci cation cannot be explained only by absolute sizes of the coe cients: Their relative sizes compared with those of other tax rules are important. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 16 / 31

17 Policy Experiments (3) Investment responses under speci cation 4 (φ tkb = 0.2, φ tcb = φ tdb = 0.01) show similar patterns to those under baseline. The responses under speci cation 2 (φ tcb = 0.2, φ tdb = φ tkb = 0.01) and speci cation 3 (φ tdb = 0.2, φ tcb = φ tkb = 0.01) show similar patterns to those under FMS speci cation Specification 1 Specification 2 Specification 3 Specification 4 Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 17 / 31

18 Policy Experiments (4) Again, Ricardian consumption shows a similar pattern to investment. Recall that initial output increase after a government spending shock in general equilibrium models is brought about by a labor hour increase and following an investment rise Specification 1 Specification 2 Specification 3 Specification 4 Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 18 / 31

19 Policy Experiments (5) Both consumption and labor income taxes have labor-dampening e ects. In fact, labor hour increases are prevented to a larger extent under speci cations 2 and Specification Specification 2 1 Specification 3 Specification 4 Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 19 / 31

20 Fiscal Multipliers under Di erent Non-Ricardian Shares SP4 SP3 SP2 Qrts ω = 0.0 ω = 0.1 ω = 0.2 ω = 0.3 ω = 0.3 ω = 0.3 by bc bi Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 20 / 31

21 Tax Rule Combination vs. Non-Ricardian Share The model under speci cation 4 delivers the crowding-in e ect on consumption even in the case where all households are Ricardian (ω = 0.0). Speci cation 4 with ω = 0.1 exhibits larger multipliers than speci cations 1-3 with ω = 0.3. A choice of tax rule combination can alter the consequences of scal policy anticipated by the given non-ricardian share. Non-Ricardian households do NOT explain all! Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 21 / 31

22 Concluding Remarks This paper presents an estimated medium-scale DSGE model of the Japanese economy detailed in scal policy. Although there is little consensus with respect to modeling scal policy rules, we con rm that it a ects scal policy e ectiveness considerably. Therefore proper modeling of the government s nancing behavior is fairly important to assess the quantitative e ects of scal policy. Direction for future research: Proper modeling of the spending side (e.g., productive public capital) Empirical evaluation of the model (e.g., comparison with VAR models) Obtaining an appropriate capital income tax series Considering the possibility that policy rules may change over time / Dealing with the zero-interest-rate period of Japan Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 22 / 31

23 Supplementary Slides Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 23 / 31

24 Related Literature Empirical Forni, Monteforte, and Sessa (2009) (FMS) rst examine scal policy e ectiveness in an estimated DSGE model augmented with multiple distortionary tax rules, utilizing (annual) scal data of the euro area. Coenen and Straub (2005) introduce distortionary taxes to the canonical Smets and Wouters (2003) model of the euro area, but in a time-invariant manner. The model is estimated without using scal data. Theoretical Aiyagari, Christiano, and Eichenbaum (1992) and Baxter and King (1993) both show that the negative wealth e ect of scal policy increases the labor supply and accordingly investment, while decreasing consumption in a neoclassical framework. Galí, López-Salido, and Vallés (2007) introduce non-ricardian households to a simple DSGE model and show that it is possible to have the empirically-supported crowding-in e ect on consumption. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 24 / 31

25 Model Overview (Smets and Wouters (2003)) (Ricardian) households utility function is additively separable between consumption and labor. (Ricardian) households supply labor and capital, and have access to government bond market. (Ricardian) households act as wage setters in monopolistically competitive labor market. There are two types of rms: perfectly competitive nal-good rms and monopolistically competitive intermediate-good rms. Nominal pro ts for intermediate-good rms are distributed to (Ricardian) households as dividends. The monetary authority follows a Taylor-style feedback rule. Real rigidities: habit formation, investment adjustment cost, variable capital utilization Nominal rigidities: sticky price and wage á la Calvo (1983), indexation in prices and wages Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 25 / 31

26 Model Features (Firms (1)) The nal-good producing rms combine the intermediate goods using the following h bundler technology: R 1 Y t = 0 y 1 i 1+λp,t t(f ) 1+λp,t df. Each intermediate-good rm f produces its di erentiated output using a Cobb-Douglus technology: y t (f ) = ε a t k t 1 (f ) α l t (f ) 1 α Φ, where k t 1 (f ) is the e ective capital stock given by k t 1 (f ) = z t k t 1 (f ), l t (f ) is the e ective labor input bundled by an independent and perfectly h competitive employment agency, which has R 1 a technology L t = 0 L 1 i 1+λw,t 1+λw,t t(n) dn. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 26 / 31

27 Model Features (Firms (2)) Intermediate-good rms reset their prices optimally with probability 1 ξ p, otherwise adjust them according to an indexation scheme γp p t (f ) = Pt 1 P t 2 p t 1 (f ). Aggregate price law of motion is then expressed as: γp 1 λ p,t λp,t P t = 4 1 ξ p (p t (f )) λp,t + ξ Pt 1 p P t 2 p t 1 (f ) 5. p t (f ) is chosen such that maximizes: E t s=0(βξ p ) s [(p t (f ) P t+s mc t+s ) y t+s (f ) P t+s mc t+s Φ] where y t+s (f ) = p t (f ) P t+s 1+λp,t+s λ p,t+s Yt+s. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 27 / 31

28 Estimation Overview Conduct Bayesian estimation using a MCMC method. The model is log-linearized around the deterministic steady state. The DYNARE software for MATLAB is applied. The draws from the posterior distribution have been obtained by taking two parallel chains of 1,000,000 replications for Metropolis-Hastings algorithm. Use Japanese data for the period 1980:Q1 to 1998:Q4. The zero-interest-rate period is not included. Use 10 data series: Other than the ordinary seven series (output, consumption, investment, labor hour, wage, in ation rate, and interest rate), I utilize government spending (as a sum of government consumption and investment) and aggregate e ective tax rates on consumption and labor income. Only quarterly data are utilized for the tax rate calculation, which is available from the National Accounts of Japan from 1980:Q1. Estimate 25 parameters assuming 12 stochastic shocks. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 28 / 31

29 Estimation Results (Structural parameter means) Euro Area U.S. Japan SW CS FMS LOWW INW SU This paper h σ c (1.00) σ l /ς (6.319) ϕ n.a ψ ξ w n.a ξ p n.a γ w n.a γ p n.a ω n.a n.a. n.a. n.a Estimates for the speci cation in which distortionary taxation is considered. Values in parentheses are calibrated. ς 1/S 00 (1), ϕ 1 + Φ/Ȳ, ψ Ψ 0 (1)/ Ψ 00 (1). Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 29 / 31

30 Estimation Results (Policy parameter means) Euro Area U.S. Japan SW CS FMS LOWW INW SU This paper ρ r φ r π φ ry ρ g n.a φ gy n.a. n.a. n.a. n.a. n.a. n.a ρ tc n.a. n.a n.a. n.a. n.a φ tcb n.a. n.a n.a. n.a. n.a ρ td n.a. n.a n.a. n.a. n.a φ tdb n.a. n.a n.a. n.a. n.a ρ tk n.a. n.a n.a. n.a. n.a φ tkb n.a. n.a n.a. n.a. n.a Estimates for the speci cation in which distortionary taxation is considered. As regards monetary policy rule, feedback from the current changes in in ation and output gap are not considered. Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 30 / 31

31 Tax Rule Speci cations Baseline FMS spec. SP1 SP2 SP3 SP4 ˆτ c rule 1 ρ tc φ tcb coe ˆτ d rule 1 ρ td φ tdb coe ˆτ k rule 1 ρ tk φ tkb coe Yasuharu Iwata (ESRI) Japanese Fiscal Policy in a DSGE Model 06/09 31 / 31