Transcript Slide 1

The Zero Lower Bound, ECB Interest
Rate Policy and the Financial Crisis
Stefan Gerlach, IMFS, and John Lewis, DNB
Introduction
• CBs across the world responded to the financial
crisis by cutting interest rates rapidly.
• Two factors may have played a role:
1. Sharp deterioration of macro economic conditions.
2. Non-zero probability that the Zero Lower Bound
(ZLB) would become a constraint.
• Did the ZLB influence ECB’s interest rate
setting?
– Difficult know & competing explanations possible.
– We argue that it probably did.
© Stefan Gerlach
2
© Stefan Gerlach
3
© Stefan Gerlach
4
ZLB and Monetary Policy
• ZLB rediscovered by Summers (1991) but seen
as a curiosity of little practical relevance to MP.
• The global decline in inflation in the late 1990s
triggered much research on the ZLB.
– Experiences of Japan.
© Stefan Gerlach
5
• One can imagine three approaches to monetary
policy in the vicinity of the ZLB.
– Let i*= f(π, y, …) denote the “optimal” interest rate
in the absence of the ZLB.
– Normally, when i* >> 0, the CB’s policy problem is
to determine i* and then sets i = i*.
– When i* < 0, the ZLB binds.
– Question is how does the CB set i as i* approaches 0?
• First approach:
– Set i = i* until the ZLB is reached and then set i = 0.
© Stefan Gerlach
6
• Second approach:
– Cut interest rates aggressively if the economy
deteriorates and maintain them at this level longer
(“extended period of time”) than implied by i*.
•
i < i* as the economy weakens.
– Reifschneider and Williams (JMCB, 2000):
•
•
•
AD determined by long interest rates, which depend on the
expected future path of short rates, i.
By setting i < i* before and after the ZLB binds, CB might
compensate for the fact i > i* when the ZLB binds.
May be possible to achieve a long interest rate similar to
that would have been observed if the ZLB had been
irrelevant.
© Stefan Gerlach
7
3. Third approach:
–
Keep the gun power dry.
•
–
–
Set i > i* so as to have more room to cut if needed.
No formal model.
Bini-Smaghi (2008):
•
•
•
Could worsen market sentiment.
If rates are cut early, little room to cut rates if economy
weakens further.
Seems to disregard the fact that on “early” cut in i makes it
less likely that i* turns negative.
© Stefan Gerlach
8
Interest rates
i
i
Time
i*
Our empirical approach
• Unfortunately, i* is not observed.
– Estimate reaction function (RF) that may shift
during sample.
•
•
•
Use pre-crisis function to predict i during crisis and think of
this as an estimate of i*.
Involves predicting i in conditions very different from those
in sample period.
Gives us a sense of when, how rapidly and why the shift
occurred.
– Compare actual i with predicted i*.
•
Did ECB cut rates faster than implied by pre-crisis RF?
© Stefan Gerlach
10
1. Switching as a function of time
• Many ways to model the shift:
– Piece-wise linear.
• Assumes break instantaneous.
– Markow switching.
• Assenmacher-Wesche (EER 2006).
– Smooth transition.
• Mankiw, Miron and Weil (AER 1987).
© Stefan Gerlach
11
Target level of interest rate:
itT     y yt    t   t    t
Gradual adjustment:


it  it 1  0 itT  it 1  1it 1  et
Reaction function:
it  ~  ~y yt  ~  t  ~ t  ~  t  ~iit 1  1it 1  et
it  Zt  et
© Stefan Gerlach
12
• Data choice:
– Overnight rate rather than repo rate!
• ON rates fell much below repo rate.
• Reflects a policy choice, not an accident.
– PMI rather than GAP.
• Available with minimal lag.
• Strongly correlated with y/y growth rate of real GDP.
– HICP inflation.
– M3.
– Nominal effective exchange rate.
© Stefan Gerlach
13
5
5
4
4
3
3
2
Repo rate
Overnight rate
2
1
1
0
07M01
0
07M07
08M01
08M07
© Stefan Gerlach
09M01
09M07
14
.06
.04
60
PMI
.02
50
.00
-.02
40
PMI
Real GDP growth
-.04
30
-.06
97
98
99
00
01
02
03
04
© Stefan Gerlach
05
06
07
08
Real GDP growth(y/y)
70
09
15
Equations for each regime:
it  I Zt  et
et ~ N 0, I2 
it  II Zt  et
et ~ N 0, II2


Composite equation:
it  1  t I Zt  tII Zt  et
Variance of errors:
  1 t   I2  t2 II2
2
2
Logistic transition:
exp  t   
t  L , , t  
1  exp  t   
© Stefan Gerlach
16
1.0
1.0
0.8
0.8
0.6
0.6
12 months/k = 0.18
6 months/k = 0.37
3 months/k = 0.73
0.4
0.4
0.2
0.2
0.0
0.0
100
10
20
30
40
50
60
© Stefan Gerlach
70
80
90
17
© Stefan Gerlach
18
© Stefan Gerlach
19
© Stefan Gerlach
20
Figure 4
1.0
1.0
0.8
0.8
0.6
0.6
Median
0.4
0.4
0.2
0.2
0.0
07M01
0.0
07M07
08M01
08M07
© Stefan Gerlach
09M01
09M07
21
Figure 5
5
5
4
4
3
3
Overnight rate
Median
2
2
1
1
0
07M01
0
07M07
08M01
08M07
09M01
09M07
One-step-ahead (static) forecasts, conditional on estimated switch.
© Stefan Gerlach
22
Figure 6
5
5
4
4
3
3
Overnight rate
Median
2
2
1
1
0
07M01
0
07M07
08M01
08M07
09M01
09M07
Dynamic forecasts, conditional on realised values of regressors and
assuming no switch.
© Stefan Gerlach
23
Summary of 1st set of results
• Evidence that the reaction function shifted.
• Interest rates were much below those predicted
by the pre-crisis reaction function.
• Compatible with the idea that ECB worried
about ZLB.
• Problems:
– No explanation for shift; only estimates of when it
and how fast it occurred.
– Return to pre-crisis reaction function not possible.
© Stefan Gerlach
24
2. Switching and economic conditions
• Allow for switch as a function of state of the
economy.
– Real GDP growth over 12 months, g.
– Interpolated.
expg t  g 
t  L, g t , g  
1  expg t  g 
© Stefan Gerlach
25
© Stefan Gerlach
26
Figure 7
1.0
1.0
0.8
0.8
0.6
0.6
Median
0.4
0.4
0.2
0.2
0.0
07M01
0.0
07M07
08M01
08M07
© Stefan Gerlach
09M01
09M07
27
Figure 8
5
5
4
4
3
3
Overnight rate
Median
2
2
1
1
0
07M01
0
07M07
08M01
08M07
© Stefan Gerlach
09M01
09M07
28
Figure 9
5
5
4
4
3
3
2
Overnight rate
Median
2
1
1
0
07M01
0
07M07
08M01
08M07
© Stefan Gerlach
09M01
09M07
29
Conclusions
• The ECB’s RF shifted during the financial crisis
at around the time of the collapse of Lehmann.
– Real economic activity drove change.
• Out finding are compatible with ZLB literature.
– Dynamic forecasts point small probability of i* < 0.
• Competing explanations possible:
– Orphanides (2010) suggests that a RF for the ECB that
uses forecasts as RHS is stable and predicts interest
rate setting also during the crisis.
© Stefan Gerlach
30