Escaping the Great Recession

Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Escaping the Great Recession1
Francesco Bianchi
Cornell and Duke
CEPR and NBER
Leonardo Melosi
FRB Chicago
Next Steps for the Fiscal Theory of the Price Level
Becker Friedman Institute
1 The
views in this paper are solely the responsibility of the authors and
should not be interpreted as re‡ecting the views of the Federal Reserve Bank of
Chicago or any other person associated with the Federal Reserve System.
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
The Great Recession and Policy Interventions
The recent recession has induced:
1. Signi…cant changes in the conduct of monetary policy, with
interest rates stuck at the zero lower bound
Standard new-Keynesian model would predict de‡ation (Bob
Hall’s puzzle)
2. A debate on the best way to mitigate the consequences of a
recession when at the zero-lower-bound:
Robust …scal intervention combined with a reduction in the
focus on in‡ation
Reluctance to explicitly abandon macroeconomic policies that
have been successful in the past
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Model Setup
We model an economy in which:
1. recurrent large negative demand shocks can force the
economy to the zero lower bound
2. two policy combinations characterize policy makers’behavior:
Monetary led policy mix: The …scal authority strongly reacts to
debt and the monetary policy rule satis…es the Taylor principle
Fiscally led policy mix: The …scal authority disregards the level
of debt and the Taylor principle does not hold
Agents are aware of the possibility of...
1. ...zero lower bound episodes,
2. ...changes in policy makers’behavior,
3. ...and the link between the two
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Main Results
1. The model accounts for the absence of de‡ation during the
Great Recession as a result of policy uncertainty about how
the rising stock of public debt will be stabilized
2. At the zero lower bound a policy trade-o¤ arises...
Announcing that …scal discipline will be abandoned greatly
mitigates the recession, but...
...it also jeopardizes long-run macroeconomic stability
3. Policymakers could escape the Great Recession by committing
to in‡ating away only the amount of debt that results from
the recession itself
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Private sector
The representative household...
1. ...maximizes expected utility
2. ...is subject to a discrete preference shock d ξ d (high or low).
t
The shock follows a two-state Markov-switching process with
transition matrix H d
The representative …rm faces...
1. ...a downward sloping demand curve
2. ...price stickiness
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Government: Monetary/…scal policy mix
Monetary rule (linearized):
"
h
i (1 ρ ) ψ p π
bt
R
π,ξ t e t + ψy ,ξ pt [y
et = 1 Z d
R
ξt
e
ρR ,ξ pt Rt 1 + σR eR ,t
+Zξ d [Zero Lower Bound]
ybt ] +
#
t
Fiscal rule:
e
τt
τ t = ρτ,ξ pt e
1+ 1
ρτ,ξ pt
h
i
δb,ξ pt e
btm 1 + ... + στ eτ,t , eτ,t
Government budget constraint + (simpli…ed) …scal rule:
e
btm
=
β
1 em
bt 1
+ bm β
e
τ t + spending
! e
btm = β
1
1
btm 1,t
R
et
π
δb,ξ pt e
btm 1 + ...
growth
N (0, 1)
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
In and out of the zero lower bound
1. Out of the zero lower bound, two alternative policy regimes:
Monetary led policy mix (AM/PF , Zξ d = 0):
t
ψπ ξ pt = M; ξ dt = h
= ψπ,M > 1
δb ξ pt = M; ξ dt = h
= δb,M > β
1
1
1
1
Fiscally led policy mix (PM/AF , Zξ d = 0):
t
ξ pt
= F ; ξ dt = h
= ψπ,F < 1
δb ξ pt = F ; ξ dt = h
= δb,F < β
ψπ
2. Negative preference shock ! Zero lower bound policy mix :
Zξ d = 1 !
t
δb ξ pt = Z ; ξ dt = l
=
Rt ! ψz R
δb,Z = 0
1
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Evolution of Policy Regimes In and Out the ZLB
Out of the zero lower bound, policymakers’behavior evolves
according to
Hp =
pMM
1 pMM
1
pFF
pFF
Combine H d with H p to obtain evolution of policy regimes in
and out the ZLB:
2
3
pMZ
p
p
H
1
p
(
)
hh
ll
5
1 pMZ
H=4
1
p
1,
1
p
(
]
hh ) [
ll
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Estimation
We solve the model with the method proposed by Farmer,
Waggoner, and Zha (2009):
St = C (ξ t , θ, H ) + T (ξ t , θ, H ) St
1
+ R (ξ t , θ, H ) εt
Agents take into account the possibility of regime changes )
Their beliefs matter for the solution of the model.
We estimate the model with Bayesian methods over the
period 1954:Q4-2014:Q1.
Regime sequence based on MS-VAR evidence (
consistent with Bianchi and Ilut (2016):
1. 1960s - 1970s ! Fiscally led policy mix
2. 1980s - 2000s ! Monetary led policy mix
3. Post-2008:Q4 ! Zero lower bound policy mix
details
) and
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Prior and Posterior Moments
ψπ,M
ψy ,M
δb,M
ψπ,F
ψy ,F
δb,F
dl
phh
pll
pMM
pFF
pMZ
Mean
1.6019
0.5065
0.0712
0.6356
0.2709
0
0.3662
0.9995
0.9306
0.9923
0.9923
0.9225
5%
1.1758
0.2980
0.0457
0.5007
0.2005
0.4827
0.9984
0.8936
0.9872
0.9888
0.8108
95%
2.0207
0.7688
0.1041
0.7546
0.3458
0.2789
0.9999
0.9599
0.9965
0.9951
0.9861
Type
N
G
G
G
G
F
N
D
D
D
D
D
Mean
2.5
0.4
0.07
0.8
0.15
0
0.3
0.96
0.83
0.96
0.96
0.50
Std
0.3
0.2
0.02
0.3
0.1
0
0.1
0.03
0.10
0.03
0.03
0.22
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Dynamics at the ZLB
GDP growth
Inflation
1
0
0.5
-1
0
-2
-0.5
2007 2008 2009 2010 2011 2012 2013 2014
2007 2008 2009 2010 2011 2012 2013 2014
FFR
1.2
1
0.8
Debt-to-GDP
Actual data
Median
90% Error Bands
0.6
300
250
200
0.4
150
0.2
2007 2008 2009 2010 2011 2012 2013 2014
2007 2008 2009 2010 2011 2012 2013 2014
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
No Policy Uncertainty
Let us consider a counterfactual that removes policy uncertainty:
Only Monetary led regime out of the ZLB
GDP growth
Inflation
0
0
Debt-to-GDP
1400
1200
-5
-5
1000
-10
800
600
-10
-15
400
Benchm ark
Only M onetary led
-20
200
-15
2007 2008 2009 2010 2011 2012 2013 2014
2007 2008 2009 2010 2011 2012 2013 2014
2007 2008 2009 2010 2011 2012 2013 2014
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
In‡ation Expectations
Corroborating evidence: In‡ation expectations
One-year horizon
Five-year horizon
Michigan survey
Median
90% error bands
1.5
1.5
1
1
0.5
0.5
0
0
2007
2008
2009
2010
2011
2012
2013
2014
2007
2008
2009
2010
2011
2012
2013
2014
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
No …scal block: Explanatory power of discrete shock
Suppose we estimate a nested model without the …scal block:
! It cannot account for the joint dynamics of growth and in‡ation
with a single shock
GDP growth
Inflation
1
0.5
Benchmark Model
Actual data
Median
90% Error Bands
0
0.5
-0.5
-1
0
-1.5
-2
-0.5
-2.5
2007
Table
2008
2009
2010
2011
2012
2013
2014
2007
2008
2009
2010
2011
2012
2013
2014
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
No …scal block: Ability to match in‡ation expectations
Suppose we estimate a nested model without the …scal block:
! It cannot account for the behavior of in‡ation expectations
One-year horizon
Five-year horizon
1.5
1.5
1
1
0.5
0.5
0
Benchmark Model
Michigan survey
Median
90% error bands
0
2007
Table
2008
2009
2010
2011
2012
2013
2014
2007
2008
2009
2010
2011
2012
2013
2014
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
In‡ation and Output gap
Inflation
2.5
2
1.5
1
0.5
0
1960
1970
1980
1990
2000
2010
1990
2000
2010
Output gap
0
-5
Benchmark
No Fiscal Uncertainty
-10
1960
1970
1980
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Policy Trade-O¤
Why do not policy makers simply announce a switch to the …scally
led regime?
Announcing the Fiscally led regime...
1. ...mitigates the recession...
2. ...at the cost of an increase in macroeconomic uncertainty
The two results are the two sides of the same coin
Announcement works only if it modi…es long term
expectations about future policy makers’behavior
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Escaping the Great Recession
Suppose policy makers commit to in‡ating away only the amount
of debt resulting from the recession itself ) No large recession
Output gap
Inflation
0
1
0.5
-5
0
-10
-0.5
10
20
30
40
50
60
10
20
FFR
1
60
40
50
60
350
0.8
300
0.6
250
0.4
200
0.2
150
30
50
400
Escaping rule
Benchmark
20
40
Debt-to-GDP
1.2
10
30
40
50
60
10
20
30
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Escaping the Great Recession
Policy makers follow the Monetary led policy mix in response to all
other shocks ) No increase in uncertainty
Output gap
Inflation
One-quarter
1.4
0.55
0.5
0.45
0.4
0.35
1.2
1
One-year
10
20
30
40
50
60
4
0.8
3
0.6
2
20
30
40
50
60
10
20
30
40
50
60
0.4
10
20
30
40
50
60
6
Two-year
10
0.8
4
0.6
2
0.4
10
20
30
40
50
60
Escaping rule
Benchmark
10
20
30
40
50
60
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Concluding Remarks
1. Recurrent ZLB events in a standard DSGE model
2. Absence of de‡ation explained by policy uncertainty in
response to a single shock
3. The model highlights a policy trade o¤ that can explain why
policy makers...
are reluctant to abandon …scal discipline
might be tempted to do so to escape the Great Recession
4. In‡ating away only the amount of debt accumulated because
of the recession would resolve the trade o¤
Introduction
The Model
Estimation
EXTRAS
Trade-O¤ and Resolution
Conclusions
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Motivating evidence
Back
We use a MS-VAR to establish a series of stylized facts:
Zt
Φξ Φ
t
= cξ Φ + Aξ Φ ,1 Zt 1 + Aξ Φ ,2 Zt 2 + Σ1/2
ωt
ξ tΣ
t
t
t
h
i
= cξ Φ , Aξ Φ ,1 , Aξ Φ ,2 , ω t N (0, I )
t
t
t
The model allows for three regimes for the VAR coe¢ cients
and three regimes for the covariance matrix
We include four observables:
1.
2.
3.
4.
Primary de…cit-to-GDP ratio
GDP growth
In‡ation
Federal Funds Rate
Conclusions
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Regime probabilities
Back
Conclusions
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Properties of the ’1960-’1970s Vs. ’1980s-2000s
De…cit/Debt
GDP Growth
In‡ation
Interest Rate
Real Int. Rate
Conditional
Regime 1
Median 16%
2.99
1.44
3.45
1.91
5.11
3.45
6.15
4.87
1.07
0.23
Conclusions
Back
Steady-States
84%
6.21
4.25
8.61
8.89
1.48
Median
3.63
3.00
2.48
4.68
2.21
Regime 2
16%
8.72
2.69
2.06
3.53
1.27
Stylized fact 1:
1960 - 1970s: High in‡ation, …scal de…cits, low real interest
rates ! Consistent with a …scally led regime
1980 - 2000s: Low in‡ation, …scal surpluses, higher real
interest rates ! Consistent with a monetary led regime
84%
0.49
3.28
2.81
5.83
3.21
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Properties of the zero lower bound regime
Deficit/Debt
20
Back
GDP Growth
4
15
Conclusions
2
10
0
5
-2
0
-4
2009
2010
2011
2012
2013
2014
2009
2010
Inflation
2011
2012
2013
2014
FFR
2
1.5
1.5
1
1
0.5
70% Bands
Discrete Shocks
Data
0
0.5
-0.5
0
2009
2010
2011
2012
2013
2014
2009
2010
2011
2012
2013
2014
Stylized fact 2: One single event is able to account for the zero
lower bound dynamics
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Pre-2008:Q3
Fiscal shocks
Deficit/Debt
10
Post-2008:Q 3
10
0.5
0
0.5
-0.2
0
10
20
20
20
0
0.5
-0.5
0
10
20
0
10
20
0
10
20
-1
0
-1.5
0
10
20
0
1
0.2
0.5
0
0
0
20
10
1
1
10
0
0
2
0
FFR
1
3
2
1
0
0
Inflation
0.2
20
4
2
0
-2
-4
Back
1
0
0
2008:Q3
GDP growth
2
1
0
-1
Conclusions
10
20
0.1
0
-0.1
-0.2
0
10
20
0
10
20
Stylized fact 3: Fiscal imbalances are important at the zero lower
bound
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Desirable properties
Desirable properties of a model aiming to explain the zero lower
bound dynamics:
1. A single large initial shock should account for the dynamics of
macro aggregates
2. The model should distinguish three periods in US economic
history
3. The model should capture the in‡ationary consequences of
…scal imbalances during the zero lower bound
Back
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Comparison: Zero lower bound dynamics
Variable
GDP growth
In‡ation
Variable
GDP growth
In‡ation
Back
Benchmark Model
Median
5%
95%
16%
0.2997 0.2324 0.4172 0.2552
0.0638 0.0312 0.1897 0.0351
Model without Fiscal Block
Median
5%
95%
16%
0.4134 0.3607 0.4723 0.3807
0.1042 0.0331 0.3075 0.0461
84%
0.3595
0.1338
84%
0.4485
0.2066
Conclusions
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Comparison: Ability to match in‡ation expectations
Benchmark Model
Variable
Median
5%
95%
Whole sample: 1-year 0.0569 0.0374 0.0857
Whole sample: 5-year 0.0451 0.0345 0.0641
0.0424 0.0317 0.0615
Pre-ZLB: 1-year
Pre-ZLB: 5-year
0.0493 0.0383 0.0734
Post-ZLB: 1-year
0.1001 0.0368 0.2067
Post-ZLB: 5-year
0.0181 0.0037 0.0958
Model without Fiscal Block
Variable
Median
5%
95%
Whole sample: 1-year 0.1675 0.1399 0.1955
Whole sample: 5-year 0.1563 0.1106 0.2141
0.1496 0.1241 0.1880
Pre-ZLB: 1-year
Pre-ZLB: 5-year
0.1836 0.1281 0.2557
Post-ZLB: 1-year
0.2309 0.1005 0.3008
Post-ZLB: 5-year
0.0656 0.0139 0.1369
Back
16%
0.0440
0.0378
0.0349
0.0415
0.0551
0.0066
84%
0.0733
0.0557
0.0529
0.0625
0.1595
0.0532
16%
0.1510
0.1307
0.1349
0.1510
0.1577
0.0298
84%
0.1840
0.1894
0.1696
0.2229
0.2778
0.1058
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Dynamics at the ZLB
Impulse response to a discrete negative preference shock d l
We consider the economy as it was in 2008Q3
A negative discrete preference shock occurs in 2008:Q4
Objectives:
1. A stylized NK model can replicate the key post-2008Q3
macroeconomic facts as a result of only one shock
2. The impulse response is not invariant with respect to the state
of the economy; in particular the …scal situation
b/c the negative preference shock implies a change in
expectations about future policymakers’behavior
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Primary de…cit and debt
The government budget constraint is given by:
btm = btm 1 Rtm 1,t / (Πt Yt /Yt
1 ) + pdt
where btm = (Ptm Btm ) / (Pt Yt ) and all variables are expressed
as a faction of GDP
In steady state:
bm
bm
bm
Also:
= (bm R m ) / (ΠM ) + pd
pd
=
1 (1 + r ) / (1 + γ )
pd
> 0 if pd < 0
=
1 1/β
pd
=1
bm
1/β < 0
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Shock to government spending
Inflation
Output gap
FFR
0.3
0.06
0.05
0.25
0.05
Monetary led
Fiscally led
ZLB
0.04
0.2
0.04
0.03
0.15
0.03
0.02
0.1
0.02
0.05
0.01
0.01
0
0
5
10
15
20
5
10
15
20
0
5
10
15
20
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Monetary/Fiscal Policy Mix
Leeper (1991) shows that two determinacy regions exist:
ψπ,ξ pt
Active Monetary, Passive Fiscal
Passive Monetary, Active Fiscal
>1
<1
δb,ξ pt
>β
<β
1
1
AM/PF ! Taylor principle is satis…ed, …scal policy
accommodates behavior of monetary authority
! Macroeconomy is insulated (Ricardian regime)
1
1
PM/AF ! Taylor principle is not satis…ed, in‡ation is free to
move to keep debt on a stable path
! Macroeconomy is not insulated (non-Ricardian regime)
Introduction
The Model
Estimation
Trade-O¤ and Resolution
Conclusions
Why does this approach work?
Policy makers are in‡uencing agents’beliefs about their long
run behavior in response to a speci…c shock
Automatic stabilizer: This behavior determines an increase in
short run expected in‡ation exactly when necessary
Policy makers are committed to raise taxes to repay the
preexisting amount of debt and all future …scal imbalances
Macroeconomic stability is retained after the recession