Download The Radical Implications of Stable, Quiet Inflation

Michelson-Morley, Occam, and Fisher:
The Radical Implications of Stable, Quiet
Inflation at the Zero Bound
John H. Cochrane
Hoover Institution, Stanford University
April 30, 2017
1 / 27
Michelson-Morley; The long quiet ZLB
6
5
10Y Govt
4
3
Core CPI
2
1
Fed Funds
0
1996
I
1998
2000
2002
2004
2006
Reserves
2008
2010
2012
2014
2016
What happens at the ZLB? Nothing.
2 / 27
Core Monetary Doctrines / ZLB predictions
What happens at ZLB?
I
Old K / adaptive E: ZLB → Deflation spiral.
I
(Friedman 68) i peg, or ZLB, or passive, is unstable.
πt+1 = (λ > 1)πt + shocks.
I
I
Taylor φ > 1 stabilizes. No Taylor, → spiral.
NK/Rational E: ZLB → π is volatile; “Self-confirming
fluctuations,” “sunspots.”
I
ZLB, peg, passive is stable but indeterminate.
Et πt+1 = (λ ≤ 1)πt ; πt+1 = Et πt+1 + δt+1 .
I
I
I
Taylor φ > 1 makes unstable, hence determinate.
φ < 1 volatility a core prediction.
MV=PY: ZLB, i ≈ 0 is irrelevant. M $50b →
$3,000b means hyperinflation. Velocity is “stable.”
QE “injects liquidity.”
3 / 27
Adaptive/Old-Keynesian Spiral
2
i
0
vr
Percent
−2
−4
π
−6
−8
−10
−12
−2
0
2
4
6
8
10
Time
xt = −σ(it − πt−1 − vtr ); πt = πt−1 + κxt ; it = max[i ∗ + φ(πt − π ∗ ), 0]
4 / 27
Core Monetary Doctrines / ZLB predictions
I
Old K / adaptive E: ZLB → Deflation spiral.
I
(Friedman 68) i peg, or ZLB, or passive is unstable.
πt+1 = (λ > 1)πt + shocks.
I
I
Taylor φ > 1 stabilizes.
NK/Rational Ex.:
I
ZLB → π is stable, but indeterminate hence volatile;
Et πt+1 = (λ ≤ 1)πt ; πt+1 = Et πt+1 + δt+1 .
I
I
I
I
At ZLB, model only pins down expected π.
Unexpected π can be anything. “Sunspots.”
φ > 1 makes π unstable, hence determinate.
φ < 1 volatility a core prediction.
MV=PY: ZLB, i ≈ 0 is irrelevant. M $50b →
$3,000b means hyperinflation. V is “stable.”
5 / 27
Michelson-Morley; The long quiet ZLB
6
5
10Y Govt
4
3
Core CPI
2
1
Fed Funds
0
1996
I
I
1998
2000
2002
2004
2006
Reserves
2008
2010
2012
2014
2016
Quiet, stable π at long period of i ≈ 0, φ << 1, huge M.
No deflation spiral. No M/QE inflation. No sunspot volatility. No
change in π dynamics. σ(π) lower?
6 / 27
US unemployment and GDP
10
Unemployment
8
6
4
Fed Funds
2
0
GDP Growth
-2
1994
I
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
Larger shock but same dynamics. Faster decline in u, lower σ(∆Y )?
E (∆Y ) is too low, but is that monetary policy?
7 / 27
Japan
6
5
4
Percent
3
10Y Govt
2
1
Int rate
0
-1
Core CPI
-2
1992
I
I
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
2014
2016
20+ years at i ≈ 0 with no spiral, sunspot σ(π).
Spiral fear understandable in 2001.
8 / 27
Europe
5
1 mo Euro Libor
4
Percent
3
2
Euro Core CPI
1
0
-1
2000
I
2002
2004
2006
2008
2010
2012
2014
2016
Lower rates ↔ lower inflation.
9 / 27
Michelson-Morley
Michelson-Morley. Experiment:
I
Inflation can be stable, quiet, at ZLB, φ < 1. Even a peg.
I
Huge excess reserves paying market interest are not inflationary.
I
φ > 1 vs. φ < 1, ZLB, is not a key state variable for σ(π), dynamics.
Implications
I
Old-Keynesian. No spiral.
I
New-Keynesian. No sunspots.
I
MV=PY. No hyperinflation.
Next theory? New Keynesian + Fiscal Theory. ...
10 / 27
NK + FTPL
∞
X
Bt−1
β j st+j
= Et
Pt
j=0
Real value of gov’t debt = PV of primary surpluses
∞
X
Bt
Pt
(Et+1 − Et )
= (Et+1 − Et )
β j st+1+j .
Pt
Pt+1
(1)
j=0
Unexpected inflation = news about pv of surpluses / debt
I
Unexpected deflation ↔ debt worth more ↔ raise tax/cut spending.
I
(1) solves spiral, indeterminacy/sunspots.
δt+1 = πt+1 − Et πt+1 ↔ fiscal policy.
I
i peg or φ < 1 can be stable (NK) and (now) determinate and quiet.
I
NK + FTPL is the only remaining pre-existing, simple, economic,
theory consistent with stable, quiet inflation at ZLB, huge reserves.
11 / 27
Occam: The (Long) Paper
What about...
I Equations? A very simple transparent model.
I Variations to rescue instability, indeterminacy, M? (A: epicycles.)
I
I
I
I
I
I
I
I
Fiscal theory objections?
I
I
I
I
I
I
I
Really unstable but QE offset deflation spiral?
NK Equilibrium selection from post-bound actions, not current φπt ?
Really active NK, no one expected it to last? (A: Japan?)
Peg still unstable/indeterminate?
Really unstable but slow to emerge (sticky wages, velocity)?
Reserves didn’t leak to M1, M2. (A: My point.)
More general models? (A: don’t change stability, determinacy.)
Large deficits, debt, Japan? (A: Low r . Not deficits, debt ↔ π.)
Previous pegs, 1970/1980, other episodes?
(A: Fiscal problems. “A peg can be stable.”)
Why is σ(π) = σ(E fiscal policy) low? (“A peg can be quiet”)
“Budget constraint,” debt repayment means passive fiscal?
(A: No; off equilibrium modeling just like NK.)
“Exogenous” surpluses? s = τ y ? s(P)? (A: No. Like dividends.)
Test FTPL? (A: Test MV=PY? P = EPV(D)?)
A: Today: I only claim NK+FTPL is possible, survives quiet ZLB
test. I do not claim it proved, explains all tests, all history.
12 / 27
Fisher
13 / 27
Frictionless model
1.5
1.5
π, with fiscal shocks
interest rate i
0.5
Percent response
Percent response
0.5
inflation π
0
π, with fiscal shocks
-0.5
inflation π
0
-0.5
-1
-1
-1.5
-3
-2
-1
0
1
2
3
-1.5
-5
-4
-3
I
I
-1
0
1
2
3
Model
πt+1 − Et πt+1
I
-2
Time
Time
I
interest rate i
1
1
it = r + Et πt+1 ,
X
= (Et+1 − Et )
β j st+j /(B/P)
“Monetary policy” changes i with no change in fiscal {s}.
Higher it raises πt+1 , immediately.
Joint fiscal-monetary tightening can give a temporary π decline.
Pricing frictions give a temporary negative π? ...
14 / 27
Effects of rate rise – Standard NK model with φ = 0
interest rate i
i
π
x
1
0.8
inflation π
0.6
Percent response
0.4
0.2
0
-0.2
output gap x
-0.4
-0.6
-0.8
-1
-4
-2
0
2
4
6
Time
I
xt = Et xt+1 − σ(it − Et πt+1 ); πt = βEt πt+1 + κxt .
I
Pricing frictions do not produce π decline.
15 / 27
Long term debt, fiscal theory, works
Simple frictionless example.
P∞
6
j=0
Pt
Price level, short term
4
(j)
(j)
Qt Bt−1
= Et
∞
X
β j st+j
j=0
2
Interest rate
I
Higher (future) i → lower
Q. P level must fall.
I
Just like a fiscal shock.
I
Then i = r + E π inflation
rises.
I
Paper: Merge with sticky
prices → smooth temporary
negative π response.
0
-2
Price level, long term
-4
-6
2
4
6
8
10
12
14
16 / 27
The Answer for negative sign?
P∞
j=0
(j)
(j)
Qt Bt−1
Pt
≈ Et
∞
X
β j st+j
j=0
Points in favor:
I
→ QE (twist), forward guidance, and i policy are the same thing.
I
Works in totally frictionless model (money, prices). (+ frictions for
realistic dynamics.)
Warnings:
I
Only works for unexpected changes. Hard to justify systematic
policy, “fine tuning.”
I
Positive in long run. Produces 1970 failed stabilizations, not
standard 1980s story. (Without a fiscal change too.)
I
Nothing like any story told to undergraduates, FOMC.
I
→ The answer is yes, but not for every question.
Other approaches?....
17 / 27
(Long) Paper: What about..
Variations that don’t work:
I
I
Sticky prices
Money U(c, M/P)
I
Only expected ∆i works. Won’t help VARs. Won’t work in IOER.
Sign helps, but off by × 10 in size.
I
Temporary rates.
I
Backward-looking Phillips, or static IS.
I
Multiple equilibria, coincident or “passive” fiscal shocks.
Active money/passive fiscal.
I
I
I
Same result with φ > 1. Solution conditional on i path (Werning). If
it = it∗ + φ(πt − πt∗ ) = iˆt + φπt produce this equilibrium observed it ,
this is πt , xt .
Standard solution of 3 equation model.
18 / 27
Paper: What about..
I
More ingredients?
I
I
I
I
Borrowing or collateral constraints, hand-to-mouth consumers,
bounded rationality or irrational behavior, a lending channel; habits,
labor/leisure, production, capital, variable capital utilization,
adjustment costs, alternative models of price stickiness;
informational, payments, monetary, financial, frictions; pricing or
timing lags, alternatives to rational expectations (“reflective,”
“k-step” expectations); non-Walrasian equilibrium, game theory,...
A: Necessary as well as sufficient. The sign (and stability?) of M
policy depends on soup, not simple economics. There is no honest
simple story to tell undergrads, FOMC.
Yes to frictions etc.! To understand size and dynamics on top of a
simple model that gets sign and stability right.
VAR evidence? (A: price puzzle, includes fiscal shocks; long term
debt effect.)
Bottom line:
I
There is no other simple, modern (rational expectations) theory,
that delivers the traditional view that higher interest rates lower
inflation, even temporarily.
19 / 27
Policy
Summary: Evidence suggests, and NK+FTPL theory digests:
I
ZLB is stable, quiet. No deflation spiral, sunspots.
I
→ Peg or passive φ < 1 too.
I
Large interest-paying reserves do not cause inflation.
I
Contrary classic doctrines were wrong.
Summary: Implication
I
I
Higher i can lead to higher π in the long run. (Neutrality.)
Negative short run effect? No simple economic model for standard
beliefs. (Only a fiscal / long-term debt channel.)
Policy: (Consequence of stability, quiet)
I
Do not fear the ZLB, balance sheet!
I
We can live the Friedman rule; Huge reserves paying market interest.
Or, better, the Treasury can issue reserves to the rest of us. No need
to keep “bonds” illiquid for price level control.
I
20 / 27
Optimal quantity of money/Balance sheet
21 / 27
Policy
Policy: (Consequence of stability, quiet)
I
The Fed can keep a low peg. (Inflation then varies as r , r ∗ vary.)
I
(Wild) The Fed can target the spread between indexed and
non-indexed debt, thus target expected inflation, and let the level of
the real rate free to respond to market forces. (Expected CPI
standard.)
it = rt + Et πt+1 → Et πt+1 = it − rt
I
The Fed can guess r , r ∗ , vary interest rates i. → More stable
inflation, output. Observe a Taylor-like rule.
I
The Fed can (try to) offset lots of shocks with time-varying
rates/spread; fine-tune inflation / output path with complex DSGE.
I
Vs. leave it alone, like hot/cold shower. Old “fine tuning,” “rules vs.
discretion,” planning vs. market debate continues.
22 / 27
Policy
The Fed? Simple rules v. fine-tuning discretion continues.
I
Observed policy may not change much – Taylorish responses to
output and inflation + temporary responses to shocks.
I
Foundations / strategy may change a lot. No more φ > 1
equilibrium selection. Fiscal anchoring. Balance sheet.
I
Monetary economics is now like regular economics! A simple S&D
benchmark, then add frictions to taste.
23 / 27
Warnings
Extrapolation warning:
I
NOT “lower rates to lower inflation” (Turkey, Brazil).
I
Must be ver persistent, credible, and with fiscal backing. (Our flight
to quality came first.)
FTPL warning:
∞
X 1
Bt−1
= Et
st+j
Pt
Rt,t+j
j=0
value of gov’t debt = pv of primary surpluses
I
Fiscal policy “anchoring” comes from expectations of eventual
primary surpluses, and low real rates for government debt.
I
Low R, flight to quality, → low P.
I
Discount rates dominate valuation everywhere.
I
Low discount rates could evaporate quickly. (Greece, but ends in
inflation.)
24 / 27
The End
25 / 27
Standard NK model with φ > 1 (Woodford)
ρ = 0.9
ρ =1
2.5
2
i π
1.5
Percent (%)
Percent (%)
2
1.5
1
0.5
0
i π
1
0.5
0
-0.5
-0.5
-1
0
2
vi
-1
4
6
8
10
0
2
vi
Time
ρ = 0.50838
4
6
8
10
8
10
Time
ρ = 0.3
0.8
0.5
0.4
Percent (%)
Percent (%)
0.6
π
0.2
i
0
-0.2
-0.4
-0.6
π
0
i
-0.5
-0.8
-1
0
2
vi
-1
4
6
8
10
0
2
vi
4
Time
6
Time
i
it = φπt + vti ; vti = ρvt−1
+ εit ; φ = 1.5
I
Standard φ > 1 model is even more Fisherian!
26 / 27
Long term debt + fiscal theory + sticky prices
P∞
(j)
j=0
(j)
Qt Bt−1
≈ Et
Pt
∞
X
j
Y
j=0
k=1
1
1 + rt+k
!
st+j ; rt = it − Et πt+1
Unexpected permanent rate rise
1
Inflation π
∆s=-4.47
Percent response
0
I
Calibrated to
2014 US
maturity
structure.
I
More sticky →
r rises, →
PV declines →
less effect.
-1
∆s=0.00
Output gap x
-2
-3
Inflation, no r effect
-4
-5
-3
-2
-1
0
1
2
3
4
5
6
7
Time
27 / 27