Optimal Fiscal Policy Lecture 1 a
Winter School 2019 Delhi School of Economics
V.V. Chari
University of Minnesota Federal Reserve Bank of Minneapolis
December 2019
Outline
Take government expenditures as given Take available tax system as given
Design taxes to maximize representative agent’s welfare Representative agent
Heterogeneous agents
Representative Agent
Ramsey Taxation
Use primal approach, develop recursive methods Findings
I No intertemporal distortions in steady state
I No intertemporal distortions with standard macro preferences
I Production efficiency is optimal
I Show multiple implementation
Ramsey Approach: Tax Instruments Given
We assume rich tax system: taxes commonly used worldwide; labor, capital, and dividend income, consumption
Private behavior modeled as competitive equilibrium Ramsey equilibrium is best competitive equilibrium Ramsey approach yields wedges
I Taxes not necessarily pinned down
I Multiple implementations
Need to take stand on initial policies or promises
I Assume value of wealth cannot be below benchmark level
Preferences and Technology
Preferences
U =
∞
X
t=0
βtu(ct,nt)
Technology
ct+gt+xt ≤F(kt,nt) (1) xt=kt+1−(1−δ)kt
ct: consumption of final good nt: labor input
gt: exogenous government consumption kt: capital
xt: investment
Benchmark Tax System
Consumption tax: τtc Labor income tax: τtn Tax on initial wealth: `0 Tax on dividends: τtd
Competitive Equilibrium: Households
max
∞
X
t=0
βtu(ct,nt) subject to
∞
X
t=0
qt
(1 +τct)ct−(1−τtn)wtnt
≤(1−`0)
"
b0+
∞
X
t=0
qt(1−τtd)dt
#
qt: intertemporal price wt: wage rate
`0: initial wealth tax b0: initial government debt dt: dividends
Competitive Equilibrium: Firms
max
∞
X
t=0
qt(1−τtd)dt
subject to
dt =F(kt,nt)−wtnt−τtk[F(kt,nt)−wtnt−δkt]−[kt+1−(1−δ)kt] dt can be interpreted either as dividends or as payments on debt Miller-Modigliani theorem applies here
Competitive Equilibrium: Government
Budget constraint
∞
X
t=0
qt
τtcct+τtnwtnt+τtddt−gt +b0
"
`0+
∞
X
t=0
qt(1−τtd)dt
#
= 0
Competitive and Ramsey Equilibrium
CE is a set of allocation{ct,nt,kt+1,dt}, prices {qt,wt}, and policies τtc, τtn, τtd, τtk, `0 such that:
I HH solves their problem
I Firms solve their problem
I Markets clear (Resource constraint holds)
Ramsey equilibrium is CE with highest utility for households
Characterization: Necessary Conditions
Lemma 1: Any CE must satisfy the implementability constraint
∞
X
t=0
βt[ctuct+ntunt] =W0, (IC) where
W0= (1−`0)uc0
1 +τ0c [(1−δ+Fk0)k0+b0] Proof:
I Substitute first order conditions of households into households’ budget constraint.
I Use first order conditions of firm to show that
∞
X
t=0
qt(1−τtd)dt = (1−τ0d)
1 + (1−τ0k)(Fk0−δ)k0
Left as exercise.
Wedges and Multiple Implementations
Key first order conditions
I Intratemporal wedge
−unt
uct = (1−τtn)Fnt
(1 +τtc) (2)
I Intertemporal wedge (1−τtd) (1 +τtc)
uct
βuc,t+1
= (1−τt+1d ) 1 +τt+1c
1 + (1−τt+1k )(Fk,t+1−δ) (3) Two of these taxes are redundant
Will say there are no intertemporal distortions if 3 is undistorted
Characterization: Sufficient Conditions
Lemma 2: Given any allocations and date zero policies that satisfy implementability constraint and resource constraint, there exist prices and policies such that the collection is a competitive equilibrium.
Proof: Setq0= 1
qt=βtuct
uc0
(1 +τc0) (1 +τct) wt =Fnt
−unt
uct
= wt(1−τtn) 1 +τtc qt(1−τtd) =qt+1(1−τt+1d )
1 + (1−τt+1k )(Fk,t+1−δ)
Characterization Theorem
Proposition 1: Any CE satisfies IC and RC. If an allocation and period 0 policies satisfy IC and RC it is implementable as a CE.
I Proof: Follows from Lemma 1 and 2.
Define a tax system as complete if it satisfies second part of proposition 1.
Incomplete Tax System and Restriction
Supposeτtc =τtd= 0 and 0≤τtk ≤1.
Then sufficiency still holds.
Need to add restriction (from intertemporal condition) that uct
βuc,t+1
−1
Fk,t+1−δ ≥0 (4)
Necessity now says if allocation satisfies IC, RC and 4 it is implementable.
Multiple Implementations
Diamond-Mirrlees system: taxes all final consumption goods, all primary inputs, here labor, and fixed factors here initial wealth: τt,τtn,`0.
I Such a tax system is always complete.
Abel system: tax dividends and labor.
I Withτ0d≤1 system may be incomplete.
Chamley-Judd system: onlyτtnandτtk.
I Ifτtk ≤1, system may be incomplete.
Wealth Restriction
Will impose wealth restriction that W0≥W¯ That is
∞
X
t=0
βt(ctuct+ntunt)≥W¯
Spirit is that the value of wealth in utility term cannot be reduced below ¯W.
Ramsey Problem
Using proposition 1 we have that Ramsey problem is to choose (ct,nt,kt+1) and period zero policies to solve
max
∞
X
t=0
βtu(ct,nt) subject to
∞
X
t=0
βt(ctuct+ntunt) =W0, where
W0= (1−`0)uc0
1 +τ0c
b0+ (1−τ0d) 1 + (1−τ0k)(Fk0−δ) k0
W0≥W¯
Characterization of Ramsey Solution
LetW =u(c,n) +λ[cuc+nun]
−Wnt Wct
=Fnt
Wct =βWc,t+1[1 + (Fk,t+1−δ)]
Hereλis the Lagrange multiplier on the resource constraint.
Proposition 2: If economy converges to a steady state, no intertemporal wedges.
Proposition 3: Ifu(c,n) = c1−σ
1−σ−φnξ+1
ξ+ 1, thenWc ∝Uc,Wn∝Un, so never intertemporal distortion, constant intratemporal distortion.
Key Property of Ramsey Allocation
Resource constraint holds with equality. That is we have production efficiency.
Production efficiency is requirement that allocation are at the boundary of the production set.
Implies intermediate goods (like investment) production decisions are not distorted.
Recursive Formulation
Consider functional equation
Vt(k,W) = maxu(c,n) +βVt+1(k0,W0) subject to
βW0+cuc+nun=W
c+gt+k0−(1−δ)k ≤F(k,n)
Ramsey problem is equivalent to functional equation ifβtVt→0.
Extension to Uncertainty
Letst be random variable
Letst be the history of shocks with probabilityµ(st) Preferences are U=P∞
t=0
P
stβtu(c(st),n(st)) Budget constraint becomes
∞
X
t=0
X
st
q(st)
(1 +τc(st))c(st)−(1−τn(st))n(st)
= (1−`0)
"
b0+
∞
X
t=0
X
st
q(st)(1−τd(st))dt
#
Resource constraint becomes
Extensions
Implementability constraint becomes X
t
X
st
βtµ(st)
c(st)uc(st) +n(st)un(st)
=(1−`0)uco
1 +τ0c
b0+ (1−τ0d) 1 + (1−τ0k)(Fk0−δ) k0
Ramsey problem is to maximize expected utility subject to IC and RC.
Can also set up problem with households making investment decisions rather than firms. Trivial to show equivalence. (See paper for details.)
Ramsey with Heterogeneous Agents
Heterogeneous Agents Model
Continuum of agents indexed byθdrawn from distribution G(θ) Allocations now indexed by type of agentθ
Preferences
u(θ) =
∞
X
t=0
βtu(ct(θ),nt(θ))
Resource constraint Z
ct(θ)dG(θ) +gt+kt+1−(1−δ)kt ≤AtF
kt, Z
θnt(θ)dG(θ)
Planner’s preferences
Z
u(θ)dλ(θ)dG(θ)
Ramsey Problem
Assume initially taxes can depend onθ Implementability constraint
∞
X
t=0
βt[ct(θ)uct(θ) +nt(θ)unt(θ)]
= (1−`0(θ)) [b0(θ) + (1 + (1−τko(θ))) (Fk0−δ)k0(θ)]
Ramsey problem
max Z
u(θ)dG(θ)dλ(θ) subject to IC and RC and LHS of IC ≥ W(θ)
Ramsey Solution
If the economy converges to a steady state, no intertemporal distortion in the steady state
If preferences are of standard macro form, no intertemporal distortion ever and constant intratemporal distortion (Same for all agents.)
That is, outcome can be implemented with constant tax rates on consumption that are the same for all agents
Restricted Tax System if Taxes Cannot Depend on Θ
Then CE must satisfy
−un(θ)
uc(θ) =θFn(θ)(1−τ) −un(ˆθ)
uc(ˆθ) = ˆθFn(ˆθ)(1−τ) Then must add restriction to characterization theorem
θuct(ct(θ),nt(θ)) unt(ct(θ),nt(θ)) =
θuˆ ct
ct(ˆθ),nt(ˆθ) unt
ct(ˆθ),nt(ˆθ) Similarly, from the intertemporal condition
uct(ct(θ),nt(θ)) βuc,t+1(ct+1(θ),nt+1(θ)) =
uct
ct(ˆθ),nt(ˆθ) βuc,t+1
ct+1(ˆθ),nt+1(ˆθ)
Recursive Formulation without Restriction
State variables arek,H(W, θ): distribution over promised wealth, andθ.
s= (W, θ).
V(k,H) = max Z
u(c(s),n(s))dλ(θ)dH(s) +βV(k0,H0)
subject to
c+g+k0−(1−δ)k =AF
k, Z
n(s)dG(θ)
βW0(θ) +uc(s)c(s0) +n(s)un(s) =W(θ)
H0(W0, θ) = Z
B(W,θ)
dH(W, θ) where
B(W0, θ) ={W;h(W, θ)≤ W0}
Recursive Formulation with Restricted Tax System
Now need to add last period’s marginal utility as a state variable s= (W,U, θ)
H is distribution over s Add restrictions
θuc(s)
un(s) =θuˆ c(ˆs)
un(ˆs) ∀θ,s,θ,ˆ sˆ and
uc(s)
U =uc(ˆs)
Uˆ s,s,ˆ U,Uˆ
Recursive Formulation with Restricted Tax System
V(k,H) = max Z
u(c(s),n(s))dλ(θ)dH(s) +βV(k0,H0)
subject to resource constraint
βW0(θ) +uc(s)c(s0) +n(s)un(s) =W(θ)
H0(W0,U0, θ) = Z
B(W,U,θ)
dH(W, θ) where
B(W0,U0, θ) ={(W,U);h(W,U, θ)≤ W0}