Monetary Policy Alternatives at the Zero Bound:
Lessons from the 1930s U.S. Christopher Hanes
March 2013
Last resorts for monetary authorities in a liquidity trap:
1) Replace inflation target with target for price level or nominal GDP
In standard NK models, credible announcement immediately boosts ∆p ,
lowers real interest rates while we are still trapped at zero bound.
“Expected inflation channel”
2) “Quantitative easing” or Large-Scale Asset Purchases (LSAPs)
Buy long-term bonds in exchange for bills or reserves
to push down on term, risk or liquidity premiums through “portfolio effects”
Can 1) work? Do portfolio effects exist?
I look at 1930s, when U.S. in liquidity trap.
1) No clear evidence for expected-inflation channel
2) Yes: evidence of portfolio effects
Expected-inflation channel: theory
New-Keynesian Phillips curve:
∆pt ' Et∆pt%1 %
Lessons from the 1930s U.S.
β
γ
( y & y n)t
T
a distant horizon T
∆pt' Et [ ∆pt%T %
β
(y & y n)t%τ ]
j
λ
τ'0
To hit price-level or $AD target, authorities must boost future (y & y n)t%τ
For any given path of y in near future, while we are still in liquidity trap,
that raises current ∆p t , reduces r t , raises y t , lifts us out of trap
Why it might fail:
- expectations not so forward-looking, rational
- promise not credible
Svensson’s “Foolproof Way” out of liquidity trap: peg to depreciated exchange rate
“a conspicuous commitment to a higher price level in the future”
Expected-inflation channel: 1930s experience
Lessons from the 1930s U.S.
Late 1920s:
International gold standard, authorities exchange currency for gold at fixed values
- fix exchange rates (within gold points)
- country with BOP deficit loses reserves, must deflate
- country with BOP surplus must inflate or buy up reserves, force losers to deflate.
“Rules of the game” say inflate
- U.S. doesn’t follow rules: instead buys up reserves (sterilizes). End game?
March 1933-January 1934: devaluation; FDR pledges “reflation”
June 1934-July 1936: unsterilized gold inflows boost high-powered money supply
July 1936-April 1937: RR8 , sterilization to prevent future inflation
April 1937-April 1938: unsterilize, Fed buys bonds
O.M.W. Sprague: “Doubtless, given time, a depreciated dollar or a devalued dollar will
(November 1933)
yield a higher price level”
Tuesday, February 26, 2013.max
Tuesday, February 26, 2013.max
Expected-inflation channel: 1930s experience
Figure 3
1
Lessons from the 1930s U.S.
6
140
120
Wholesale price indexes
Cotton
Rubber
Raw materials
100
80
60
7
40
Exchange rate,
French franc
(cents per franc,
left axis)
6
20
0
5
4
3
1933
1934
1935
1936
1937
Expected-inflation channel: what would evidence be?
Lessons from the 1930s U.S.
Nonag wages (or price index for nonag domestic value-added)
T
∆wt' Et[∆wt%T%
β
λ
T
j (y &y )t%τ] ' Et[∆wt%T %
d
n
τ'0
output gap ü
β
λ
d
(y
&λn)t%τ] %
j
τ'0
β
λ γ
µt
output deviation from trend ü
µ : desired wage mark-up, reflects workers’ bargaining power, efficiency wages etc.
Usually, E t yt%1 . ρ yt
∆wt' Et∆wt%T %
β
1
λ
1&ρ
(y d&λn)t %
β
λ γ
µ t % zt
z: expected-inflation channel
wagesetters’ forecast for cumulative future output
deviates from usual relation with current output
Expected-inflation channel: how I look for evidence
Lessons from the 1930s U.S.
1) Using postwar data, regress ∆w on real activity and lagged wage inflation
2) Apply coefficients to 1930s real activity projecting forward from January 1929
What to do with lagged-inflation coefficients? On and off.
3) Look at deviations of actual ∆w from projected path
Are deviations consistent with operation of expected-inflation channel?
Yes: ∆w anomalously high in 1933, 1934 (devaluation, inflation talk)
falls back toward projections in 1936 ( RR8 , sterilization)
up again in 1937 (desterilization, Fed buys bonds)
No: deviations easily explained by NIRA, unionization (recall µ )
Expected-inflation channel: NIRA, unionization
Lessons from the 1930s U.S.
Figure 6
June 1933: NIRA bars employers from firing strikers, union organizers
August 1933: Blanket code (President’s Re-employment Act) with minimum wages
August-December 1933: industry codes, further wage increases
July 1935: Wagner Act
November 1936: FDR re-elected, many companies give in to bargaining
300000
40
June
1933
Blanket
code
35
200000
30
150000
Strikers
Union density
25
100000
20
50000
0
15
27
28
29
30
31
32
33
34
35
36
37
38
39
percent
number of workers
250000
FDR
re-election
Expected-inflation channel: projections and actual wage inflation
Figure 8
1
0.3
I
6
7 II 9
Twelve-month change in log AHE
0.2
Actual
0.1
0.0
-0.1
Projections
With lagged-inf. effect
Without lagged-inf. effect
-0.2
-0.3
28
29
30
31
32
33
34
35
36
37
1. March 1933: Bank Holiday. Banks and U.S. Treasury cease gold payments
6. End of January 1934: Gold Reserve Act sets new gold price
I August 1933: NRA Blanket code
II November 1936: Roosevelt wins re-election
38
39
40
Expected-inflation channel: looking for evidence
Figure 9
1
I
6
Lessons from the 1930s U.S.
II
130
AHE Mfg/
AHE Services
(index number)
0.8
120
110
$/hr
Wage index,
manufacturing
0.6
100
AHE manufacturing
(left scale)
0.5
90
0.4
29
30
31
32
33
34
35
36
37
1. March 1933: Bank Holiday. Banks and U.S. Treasury cease gold payments
6. End of January 1934: Gold Reserve Act sets new gold price
I August 1933: NRA Blanket code
II November 1936: Roosevelt wins re-election
38
39
index number
0.7
Portfolio effects: theory
Lessons from the 1930s U.S.
How do LSAPs affect bond yields?
1) They don’t really, but financial market participants price in a chance they do
2) “Signalling channel”: policymakers’ intentions for future overnight rates
3) Portfolio effects
- “Available local supply,” “scarcity” or “market segmentation” channel
Effects concentrated on securities with similar characteristics
- “Duration” channel (Gagnon, Raskin, Remache, Sack 2011)
LSAPs remove long-term bonds (subject to duration or interest-rate risk) from portfolios,
replace them with zero-duration assets (bills or high-powered money).
Term premiums must fall to make investors willing to hold new portfolio.
1) and 2) operate only if operations are publicized
3) operates even if financial-market participants are unaware of operation
Duration channel can be framed in terms of money demand (following Keynes; Tobin 1958)
Portfolio effects: duration risk and money demand
Lessons from the 1930s U.S.
“Preferred-habitat” investors and “arbitrageurs” (Vayanos & Vila 2009)
Arbitrageurs hold M “money” (reserves, cash, zero-interest bills)
B bond portfolio
P unit price of bond portfolio
Et Pt%1
2
σP perceived variance of ( Pt%1&E tPt%1 )
q Var(At%1)
maximize Et Pt%1Bt % Mt &
2
At
ö mt
d
1
1
. a t&
q σP/EtPt%1
2
iˆt
(1%iˆt)2
where iˆt ' Et Pt%1/P t & 1
s.t. At ' Mt % Pt Bt
d
Miˆt/Mmt . & q σP/EtPt%1 2
Portfolio effects: 1930s natural experiments
Figure 5
7
20000
8
Lessons from the 1930s U.S.
9
10
11
$Millions
15000
High-powered
money
10000
Gold stock
5000
Nonborrowed
reserves
Treasury balances
0
4/07/34
3/07/36
2/05/38
1/06/40
Portfolio effects: 1930s natural experiments
Lessons from the 1930s U.S.
What should I observe in weekly data on bond yields?
d
md
m t ' &νiˆt % εt
“Money” (HPM, bills) demand:
Gold flow from BOP:
(
∆g t ' κ ∆ (iˆt & it &Et∆et%1) % εt& ∆forres
“Money” supply:
ms
bop
ms
∆m t ' ∆g & ∆tres % ∆εt ' κ ∆ iˆt & ∆tres % εt % ∆εt
(
& κ ∆ (it %Et∆et%1)% εt& ∆forres ü
þ
∆iˆt ' &
1
ν
∆m t %
1
ν
md
∆εt
'
1
ν
∆trest&
1
ν
1
∆gt %
ν
md
∆εt
&
1
ν
ms
∆εt
úcorrelated with ∆g , ∆m ?
∆m t '
∆g t '
ν
ν%κ
1
κ%ν
κ
(εbop % ∆εms & ∆tres)t %
ν%κ
νεbop% κ ( ∆tres & ∆εms)
%
∆iˆ has bigger effect on shorter-duration bond
t
md
∆εt
κ
ν%κ
md
∆εt
Portfolio effects: 1930s natural experiments
Lessons from the 1930s U.S.
Regression results Table 2
Yields: weekly average ending Saturday
April 1934-July 1936
Money: weekly Wednesday
117 weeks Quadratic time trends included on RHS
Treasuries
Medium-term notes
(1)
∆Ln(HPM1)
[Robust SE]
p-value
-1.016
[0.332]
0.00
(2)
-0.934
[0.315]
0.00
-2.638
[1.907]
0.17
∆g
2
(4)
-0.190
[0.146]
0.20
-3.524
[1.984]
0.08
(5)
0.03
0.03
0.03
Corporate (BAA)
(6)
-0.171
[0.140]
0.23
-0.610
[0.796]
0.45
0.966
[0.304]
0.00
∆tres
R̄
(3)
Long-term bonds
(7)
(8)
-0.801
[0.350]
0.02
-0.715
[0.326]
0.03
-0.774
[0.821]
0.35
-2.744
[2.163]
0.21
0.162
[0.139]
0.25
-0.01
-0.02
-0.02
(9)
-3.422
[2.088]
0.10
0.756
[0.337]
0.03
0.07
0.08
0.08
Portfolio effects: 1930s natural experiments
Lessons from the 1930s U.S.
Percent yield to maturity
4
3
Treasury bonds
Notes April 34 - July 36
July 36 - April 38
April 38 - August 39
Bonds April 34 - July 39
July 36 - April 38
April 38 - August 39
2
1
0
9.0
Treasury notes
9.2
9.4
9.6
Log of high-powered money
9.8