Discussion of
Liquidity Traps and Expectation Dynamics: Fiscal Stimulus or Fiscal Austerity By Jess Benhabib, George W. Evans and Seppo Honkapohja
Sebastian Schmidt (ECB)
SNB Research Conference 20 September, 2013 Disclaimer: The views expressed on the slides are those of the author and do not necessarily represent those of the ECB.
1 / 25
Introduction Under conventional monetary policy, the zero lower bound on nominal interest rates (ZLB) poses the threat of multiple equilibria and “deflationary spirals”.
This paper investigates the desirability of policies tailored to avoid disastrous outcomes in the wake of ZLB events in a New Keynesian model where agents form expectations using adaptive learning rules.
Focus on fiscal policy. 2 / 25
Framework
Small, non-linear New Keynesian model
Adaptive infinite-horizon learning
Agents have point expectations
Focus on steady state learning
3 / 25
‘Normal’ policy Suppose, monetary policy ∗is characterized by e φy e AR R∗ −1 yt π , Rt = 1 + (R∗ − 1) πt∗ y∗ and government consumption is constant gt = g. There exist two steady states, the intended steady state π = π ∗ and an unintended steady state π = πL < π ∗ . The intended steady state is locally but not globally stable under learning. The unintended steady state is unstable under learning. Threat of unstable deflationary paths. 4 / 25
Stability under learning (π ∗ = 1.02, A = 2.5, φy = 0) unstable stable 0.955
Ye
0.95
0.945
0.94
0.935
0.93 0.97
0.98
0.99
1
1.01 e
π
1.02
1.03
1.04
1.05 5 / 25
Stability under learning (π ∗ = 1.005, A = 2.5, φy = 0)
0.955
Ye
0.95
0.945
0.94
0.935
0.93 0.97
0.98
0.99
1
1.01 e
π
1.02
1.03
1.04
1.05 6 / 25
Fiscal policy
Can fiscal policy contribute to mitigate/eliminate deflationary threats? Benhabib, Evans and Honkapohja consider three policies:
Temporary pre-specified fiscal stimulus Temporary pre-specified fiscal austerity Fiscal switching rule
7 / 25
Fiscal switching rule Select some inflation threshold π ˜ > πL . If πt | (gt = g) < πte < π ˜, choose gt > g such that πt > πte , else if πte ≤ πt | (gt = g) < π ˜, ˜, choose gt > g such that πt > π otherwise choose gt = g.
8 / 25
Fiscal switching rule c’td
Not necessary to decide in advance about magnitude and duration of fiscal stimulus. Agents do not know the future path of government spending, i.e. agents do not know the fiscal rule. Eliminates unintended steady state and seems to be a robust policy against deflation traps.
9 / 25
Inflation - Output dynamics (π0e = 0.993, y0e = 0.9425, g0e = g) 1.01 1 0.99
output
0.98 0.97 0.96 0.95 0.94 0.93 0.92 0.98
1
1.02
1.04
inflation
1.06
1.08
1.1 10 / 25
Paths towards intended steady state (π0e = 0.993, y0e = 0.9425, g0e = g) output
expected output
1 0.945
0.95 0
500
1000
1500
2000
0.94
0
inflation
500
1000
1500
2000
expected inflation
1.1 1.02 1.05 1
1 0
500
1000
1500
2000
0
government consumption
500
1000
1500
2000
expected gov. con. 0.202
0.25 0.201 0.2
0
500
1000
1500
2000
0.2
0
500
1000
1500
2000 11 / 25
Comments on
1. Inflation, output dynamics
2. Fiscal policy rule
3. Treatment of real interest rate expectations
12 / 25
1. Inflation, output dynamics
We observe large swings in adjustment dynamics. Potential remedies:
Change parameter values in monetary policy rule. Enhance fiscal switching rule, allowing for gt < g if π and/or y (otherwise) above some upper threshold.
13 / 25
Inflation - Output dynamics (π0e = 0.993, y0e = 0.9425, g0e = g) 1.01 φ =50 y
φ =1500
1
y
0.99
output
0.98 0.97 0.96 0.95 0.94 0.93 0.92 0.98
1
1.02
1.04
inflation
1.06
1.08
1.1 14 / 25
2. Fiscal policy rule Fiscal rule considered in the paper requires “reverse engineering”.
Could a simple(r) fiscal rule do the job? Consider for instance: gt = g
π∗ πte
Bπ∗ g
y∗ yte
ψy
,
and assume that private agents do not know the rule, and g0e = g.
15 / 25
Stability under learning: Only monetary policy rule unstable stable 0.955
Ye
0.95
0.945
0.94
0.935
0.93 0.97
0.98
0.99
1
1.01 e
π
1.02
1.03
1.04
1.05 16 / 25
Stability under learning: Simple fiscal rule (B = 1.5, φy = ψy = 0)
0.955
Ye
0.95
0.945
0.94
0.935
0.93 0.97
0.98
0.99
1
1.01 e
π
1.02
1.03
1.04
1.05 17 / 25
Stability under learning: Simple fiscal rule (B = ψy = 3.5, φy = 50)
0.955
Ye
0.95
0.945
0.94
0.935
0.93 0.97
0.98
0.99
1
1.01 e
π
1.02
1.03
1.04
1.05 18 / 25
Paths towards intended steady state (π0e = 0.993, y0e = 0.9425, g0e = g) output
expected output
1.1 1
0.95
0.9 50 100 150 200 250 300
0.94
inflation
50 100 150 200 250 300
expected inflation 1.04
1.1
1.02
1 0.9
1 50 100 150 200 250 300
50 100 150 200 250 300
government consumption
expected gov. con. 0.3
0.2 0
0.2
50 100 150 200 250 300
0.1
50 100 150 200 250 300 19 / 25
3. Treatment of real interest rate expectations With arbitrary inflation expectations, for some regions of the space of expectations the expected gross real interest rate could become negative. Authors truncate private agents’ real rate expectations at some horizon T . Beyond T agents expect the real rate to be equal to its steady state 1/β. How does the choice of T affect stability properties and inflation-output dynamics? 20 / 25
Stability under learning: T = 28 unstable stable 0.955
Ye
0.95
0.945
0.94
0.935
0.93 0.97
0.98
0.99
1
1.01 e
π
1.02
1.03
1.04
1.05 21 / 25
Stability under learning: T → ∞
0.955
Ye
0.95
0.945
0.94
0.935
0.93 0.97
0.98
0.99
1
1.01 e
π
1.02
1.03
1.04
1.05 22 / 25
Conclusion There is not only a role for fiscal policy to improve stabilization outcomes in the face of zero lower bound events, but also to avoid desastrous outcomes.
Benhabib, Evans and Honkapohja stress the role of fiscal stabilization policy for ensuring stability under learning.
From a welfare perspective, avoiding deflationary spirals should be of utter importance.
23 / 25
Backup slides
24 / 25
Temporary equilibrium equations New Keynesian Phillips curve:
Q (πt ) = (πt − 1) πt β ν − 1 yt ν ν 1+ǫ ν − 1 yte e 1+ǫ α α y + (y ) − Q (πt ) = − γα t γ yt − gt 1 − β γα t γ yte − gte Consumption function:
e πt
yt = gt + (1/β − 1) (yte − gte )
T πte 1 − 1+f (πe ,ye ) t
t
1 + f (πte , yte ) − πte
+
βT 1/β − 1
25 / 25