Lecture 8: Fiscal Policy
Patrick Macnamara
ECON60111: Macroeconomic Analysis University of Manchester
Fall 2020
Government Budget Constraint
2/33
Dynamics of Debt
• Evolution of government debt:
D ̇ (t) = [G(t) − T(t)]+r(t)D(t) (1)
primary deficit
• Definition of terms:
• G(t) = real government purchases at time t
• T(t) = real government tax revenues at time t • D(t) = real government debt at time t
• r(t) = real interest rate at time t
• For what follows, let’s also define R(t) = t r (τ)dτ as the 0
continuously compounded interest over the period [0,t].1
1This specification accommodates the fact that r(t) may change over time. If r(t) is constant at ̄r, then R(t) = ̄rt.
3/33
Government Budget Constraint
• Start with law of motion for debt:
D ̇ (t) = [G(t) − T(t)] + r(t)D(t)
• Multiply both sides by e−R(t); integrate from t = 0 to t = s:
s −R(t) ̇ s −R(t) s −R(t) e D(t)dt = e [G(t)−T(t)]dt+ e
r(t)D(t)dt
000
• Use integration by parts to replace LHS:
s −R(t) ̇ −R(s) s
• Then we have
−R(t)
e D(t)dt = D(s)e − D(0) + e 00
r(t)D(t)dt Details
D(s)e
−R(s)
= D(0) +
s 0
e
−R(t)
[G(t) − T(t)]dt
4/33
Government Budget Constraint
• From the dynamics of debt, we have (see previous slide):
−R(s)
s 0
−R(t)
[G(t) − T(t)]dt (2)
D(s)e
• Let’s also impose no-Ponzi-game condition:
= D(0) +
e
lim e−R(s)D(s) ≤ 0 (3) s→∞
In the limit, the present value of the government debt ≤ 0. • Take limit of (2) as s → ∞, imposing (3):
∞ −R(t) ∞ −R(t) e G(t)dt ≤ −D (0) + e
T(t)dt (4)
00
PDV of gov’t purchases PDV of taxes
This is the government’s B.C.; it incorporates debt choices the government must make given the path of G(t) and T(t).
5/33
Implications
• Re-write government’s budget constraint, (4):
0
(in present value) to offset its initial debt.
• Nevertheless, the following examples can satisfy the budget constraint:
• Positive but constant value of D
(i.e., government never pays off its debt)
• D grows at a constant rate, less than the real interest rate (i.e., government debt grows without bound)
See the following slides for more details.
∞ −R(t) e
[T(t)−G(t)]dt ≥D(0) Government must run primary surpluses large enough
6/33
Example: Constant Debt
• Suppose r(t) = r > 0 (i.e., r is constant)
• Suppose D(t) = D > 0 (i.e., D is constant) • For D to be constant, we need
D ̇ =G−T+rD=0 ⇒ T−G=rD>0 Government’s primary surplus is constant and positive.
• Plug T − G = rD back into the government’s B.C.:
D(0) ≤
D ≤ rD
[T(t) − G(t)]dt ∞ −rt
∞ −rt e
0
0
e dt = D
=1/r
⇒ Government budget constraint is satisfied.
7/33
Example: Permanently Growing Debt
• Suppose r(t) = r is constant and D grows at a constant rate, with D ̇ /D = γ < r.
• For D to be growing at rate γ, we need: D ̇ G − T + rD
D = D = γ ⇒ T − G = (r − γ)D
Primary surplus must always be positive, and proportional to
the amount of outstanding debt.
• PlugT−G=(r−γ)D=(r−γ)D(0)eγt backintothe government’s B.C.:
0
∞ −rt
e [T(t) − G(t)]dt
D(0) ≤
D(0) ≤ (r − γ)D(0)
∞ 0
dt = D(0) ⇒ Government budget constraint is satisfied.
−(r−γ)t
e
=1/(r−γ)
8/33
Example: Economic Growth
• Suppose r(t) = r is constant and output Y grows at a constant rate, with Y ̇ /Y = g.
• ̇ Suppose the debt-to-GDP ratio is constant, (D/Y ) = 0
• ̇ ̇ ̇ Since(D/Y)/(D/Y)=D/D−Y/Y =0:
D ̇ Y ̇ G − T + rD
D=Y=g ⇒ D =g ⇒ T−G=(r−g)D
Debt grows at constant rate g, D/Y constant, but...
• if r > g, government must run primary surpluses that
increase with D, government B.C. satisfied
(see previous slide).
• if r < g, government can run primary deficits forever; but
no-Ponzi game condition is violated.
9/33
U.S. Federal Surplus/Deficit
5 4 3 2 1 0 -1 -2 -3 -4 -5 -6 -7 -8 -9
-10 1960
Total Surplus Primary Surplus
1970 1980
1990 2000 2010 2020 Year
Source: Congressional Budget Office (CBO)
10/33
Surplus (Pct. of GDP)
Sustainability of U.S. Fiscal Policy
• U.S. federal government has run frequent primary deficits over last 4 decades (especially since 2008)
Average primary deficit 1980–2019, 1.1% of GDP
• Will future taxes be enough to meet future promises regarding government spending – i.e., is there a fiscal imbalance?
• Auerbach (1997) measure of “fiscal gap”.
• By what constant fraction of GDP would taxes have to be
increased for the government budget constraint to be
satisfied (given projections of future taxes and spending)?
• Fiscal gap ∆ is the solution to:
0 YPROJ(t)
∞ PROJ TPROJ(t)−GPROJ(t)
e−R (t) +∆ YPROJ(t)dt=D(0)
11/33
Sustainability of U.S. Fiscal Policy
• Auerbach and Gale (2017): U.S. Fiscal Gap ranges from 5.4% to 9.3%2
• This is large. For comparison, federal tax revenues are only about 18% of GDP.
• Large projected increases in federal spending on Social Security, Medicare and Medicaid (due to demographic changes and medical progress)
• However, there are large uncertainties in these projections.
• Large adjustments needed to satisfy budget constraint.
e.g., spending reductions, tax increases, default
• For comparison purposes, consider some European fiscal
gaps:3 UK 4.3%, France 0.2%, Germany 2.2%
2Estimates differ depending on projections for future Medicare spending. 3Source: Table 4.2, EC’s Debt Sustainability Monitor 2019
12/33
Ricardian Equivalence
13/33
Ricardian Equivalence
• How should the government choose between taxes and bonds, for a given path of government spending?
• Consider the household’s budget constraint in the Ramsey-Cass-Koopmans model, with taxes:
∞ −R(t) ∞ −R(t) e C(t)dt ≤K(0)+D(0)+ e
[W(t)−T(t)]dt where D(0) is the bond holdings of the household and
00
K (0) + D(0) is its initial wealth.4
• Suppose government B.C., (4), holds with equality:
∞ −R(t) ∞ −R(t) e T(t)dt = e
00
G(t)dt +D(0) (5)
4We assume number of households H = 1 and total population L(t) = 1. 14/33
Ricardian Equivalence: Consumption
• Use government B.C., (5), to re-write household’s B.C.: ∞ −R(t) ∞ −R(t) ∞ −R(t)
e C(t)dt ≤ K(0)+ e W(t)dt− e G(t)dt 000
Household’s budget constraint depends only on the present value of government purchases.
• It’s reasonable to assume that taxes do not enter directly into household utility.
• Therefore, for a given path of government spending, G(t), path of taxes T(t) does not matter for consumption.
15/33
Ricardian Equivalence: Capital
• Household is indifferent between holdings of capital and government bonds because both offer the same return.
• Therefore, let A(t) = K(t) + D(t).
• Law of motion for total household assets, A(t):
A ̇ ( t ) = r ( t ) A ( t ) + W ( t ) − T ( t ) − C ( t )
• Using A ̇ = K ̇ + D ̇ :
r[K+D]+W−T−C=K ̇ +[G−T]+rD
A ̇ D ̇ ⇒ K ̇ = rK + W − G − C
Government purchases, not taxes, affect capital accumulation.
16/33
Ricardian Equivalence: Summary
• This result is known as Ricardian equivalence.
For a given pattern of government purchases, the timing of taxes and the method of financing does not affect consumption or capital accumulation.
• To see the logic, suppose the government lowers taxes T today, increasing household’s after tax-income W − T
• Government then issues more debt, since D ̇ = (G − T ) + rD. • Government B.C. implies future taxes must increase.
• Household increases savings (in the government bond),
A ̇ = rA + W − T − C.
• Household saves the tax cut (to pay higher future taxes),
leaving consumption and capital accumulation unaffected.
• Result differs from usual economic intuition – e.g., U.S. cut taxes in 2008, with the objective of increasing consumption.
17/33
Ricardian Equivalence in Practice
• Does Ricardian equivalence hold in practice? If it doesn’t hold, are there large departures from Ricardian equivalence?
• Turnover in population (e.g., overlapping generations model) is one reason Ricardian equivalence might not hold.
• Some of the future tax burden associated with a bond issue today are borne by individuals who are not alive yet.
• Yet, intergenerational links can cause a series of households with finite lifetimes to behave as one household with an infinite horizon.
• Households live long enough that most of the future taxes will be levied during the lifetime of individuals alive today.
• Even still, due to the permanent income hypothesis (PIH), a temporary change in after-tax income due to a tax cut today, would have a small effect on consumption.
18/33
Ricardian Equivalence and the PIH
• The issue of whether Ricardian equivalence is a good approximation depends on whether the PIH is a good description of consumption behavior.
• Consider tax cut / bond issue today, which is financed by higher taxes in future
• While path of after-tax income changes, present value of after-tax income is the same ⇒ consumption unaffected if PIH holds.
• We saw in lecture 4 on consumption that there are significant empirical deviations from the PIH
• Let’s consider two examples which cause the PIH to fail.
1. Liquidity/borrowing constraints. 2. Precautionary-savings motive.
19/33
Liquidity Constraints
• Suppose the government lowers taxes today and issues a bond to finance the tax cut.
• Suppose households either cannot borrow or borrow at a higher interest rate than the government.
• In this case, household consumption could increase.
• Government is essentially borrowing on behalf of the
household – lower taxes today in return for higher taxes in
the future.
• This benefits constrained households, who would increase
consumption.
20/33
Precautionary Savings
• Suppose the government lowers taxes today and issues a bond to finance the tax cut.
• With lump-sum taxes, the tax cut still has no effect on the household’s budget constraint.
• However, taxes usually depend on income (and are progressive)
• Distortionary income taxes generate an insurance benefit: low taxes if future income is low, higher taxes if future income is high.
• An increase in taxes in the future (but only for high income states) will increase the insurance benefit, reducing the incentive to save ⇒ consumption increases today.
21/33
Tax Smoothing
22/33
Tax Smoothing
• When Ricardian equivalence holds, the actual time path of taxes is indeterminate.
• Therefore, what actually determines the deficit?
• We’ll develop a model, due to Barro (1979), in which deficits
are chosen optimally.
• Assume there are distortions from raising revenue.
• Before, we have considered lump-sum taxes.
• However actual taxes are a function of income – hence they
distort behavior and cause inefficiencies.
• Barro assumes these distortions increase more than
proportionally with the amount of revenue raised.
• This creates an incentive for the government to smooth taxes over time.
23/33
Tax Smoothing Under Certainty
• Assume time is discrete.
• There is no uncertainty (for now).
• Path of output (Yt) and government purchases (Gt) are exogenously given.
• Interest rate (r) is exogenous and constant.
• There is some initial value of government debt, D0.
• Government will choose the path of taxes (Tt) to minimize the present value of the distortion costs created by taxes, subject to the government’s budget constraint.
24/33
Costs of Distortions
• Assume the distortion costs from raising taxes Tt are given by
Ct Tt
Y =f Y , f(0)=f′(0)=0, f′′(·)>0
tt
where Ct is the cost of distortions in period t.
• These assumptions imply:
• f′(T/Y)>0whenT/Y >0.
• Ct/Yt rises more than proportionally with Tt/Yt.
• Be careful with the notation here – following the Romer textbook, Ct is the distortion cost, not consumption!
25/33
Government’s Problem
• The government’s problem:
∞ 1 Tt min t Ytf
{Tt}∞t=0 t=0 (1+r) Yt
Ct
subject to
∞ Tt =D +∞ Gt (6)
(1+r)t 0 (1+r)t t=0
t=0
• Government minimizes present value of distortion costs
subject to its budget constraint.
• Equation (6) is the discrete-time analogue of the continuous-time budget constraint we saw earlier (see Equation (4)).5
5Equation (6) can be derived by combining individual period-t constraints, Dt+1/(1 + r) = Dt + Gt − Tt and the no-Ponzi game condition, limT→∞DT/(1+r)T =0.
26/33
Solving the Model
• Set up the Lagrangian:
∞ 1 Tt ∞ Gt−Tt
L= Yf +λD+
(1+r)t t Y 0 t=0 t
• FOC:
∂L Ytf′Tt 1 λ
t=0
(1+r)t T
=YtYt− =0⇒λ=f′t ∂Tt (1+r)t (1+r)t Yt
• This condition holds for all t. Hence:
Tt Tt+1 Tt Tt+1
f′ =f′ ⇒ = (7) Yt Yt+1 Yt Yt+1
Since f ′′(·) > 0, the tax rate is constant over time. • To find the tax rate, use (7) with the government’s B.C.
27/33
Implications
• What’s the implication of this model for the deficit?
• Letτ=T0/Y0 =T1/Y1 =T2/Y2 =···
be the constant tax rate
• The government’s primary deficit, as a percentage of GDP:
Gt − Tt = Gt − τ
Yt Yt
Government runs a primary deficit when Gt /Yt is high, and a primary surplus when Gt/Yt is low.
28/33
Adding Uncertainty
• Now assume that path of Gt and Yt are uncertain.
• The government’s problem is now
∞ 1 Tt
minE Yf
0 (1+r)t t Yt
t=0 subject to
Dt+1 =Dt +Gt −Tt forallt 1+r
• Due to uncertainty, we use the government’s period-t budget constraint rather than the infinite-horizon budget constraint.
• This budget constraint is the discrete-time analogue to the continuous-time equation we saw earlier
(D ̇ = (G − T) + rD, see Equation (1)).
• Government chooses the tax Tt in each period, after learning the value of Yt and Gt.
29/33
Solving the Model with Uncertainty
• Set up the Lagrangian, starting from time t:
L =Et
∞ 1 Tt+s
s=0
(1 + r)s Yt+sf
Y + t+s
Et
λ t + s (1+r)s
∞ s=0
D t + s + 1 Dt+s +Gt+s −Tt+s − 1+r
• FOC: ∂L
Tt Yt
Tt Yt
⇒ λt=f′
∂D =−1+r+Et 1+r =0 ⇒ λt=Et[λt+1]
∂Tt ∂L
=f′
−λt=0
λt • Therefore, we get:
f′
λt+1
t+1
Tt Yt
Tt+1
=Et f′
Yt+1
30/33
Solution with Uncertainty
• To summarize, with uncertainty, we get:
Tt+1
Tt Yt
f′
There can be no predictable changes in the marginal
distortion costs.
• Assume f (·) is a quadratic.6 Then:
Tt Tt+1
=Et f′
Yt+1
Y=Et Y (8) t t+1
There cannot be predictable changes in the tax rate ⇒ the tax rate follows a random walk.
6For example, if f (x) = a x2. 2
31/33
Implications
• Suppose D0 = 0 and suppose Gt /Yt follows a random walk:
Gt+1 Gt Et Y =Y
t+1 t
• Then a balanced-budget policy, Tt /Yt = Gt /Yt will also
follow a random walk, satisfying (8)
• In this case, the primary deficit is always zero.
• This suggests that deficits and surpluses arise when the Gt/Yt is expected to change.
32/33
Implications
• Suppose for example E Gt+1 < Gt . t Yt+1 Yt
That is, government spending is expected to fall.
• Tomorrow’s expected primary deficit (as a % of GDP):
Gt+1 Tt+1 Gt+1 Tt Gt Tt Et Y −Y =Et Y −Y
the primary deficit is expected to increase.
33/33
Appendix
1/2
Using Integration By Parts
• If u = u(t) and v = v(t), then the integration by parts formula states that
b′ bb′
u(t)v (t)dt = [u(t)v(t)]a − u (t)v(t)dt
aa
• Use u(t) = e−R(t) and v′(t) = D ̇ (t), implying
u′(t) = −r(t)e−R(t) and v(t) = D(t) • Then re-write:
s
s s
e−R(t) D ̇ (t)dt = e−R(t) D(t)
−r(t)e−R(t) D(t)dt
0
u(t)
0
− D(s) − D(0) +
′
u(t) −R(s)
v(t)
v(t) 0
u′(t) −R(t)
v(t) r(t)D(t)dt
Back
= e
s 0
e
2/2