Transcript by Ghosh, Qureshi and Tsangarides - Faculty Directory | Berkeley

Discussion of
Friedman Redux … by
Ghosh, Qureshi and Tsangarides
Andrew K. Rose
Berkeley-Haas, NBER and CEPR
A Critique of a Critique
• “… no strong, robust or monotonic relationship
between exchange rate regime flexibility and the rate
of current account reversion …”
– Chinn-Wei
• Response here necessarily involves overturning
negative finding with strong robust relationship
– Trick: use bilateral (not multilateral) relationships
– Ex: US vs. China AND vs. Canada AND vs. Mexico …
• Not US vs. RoW
• Gratuitous personal reference: Rose and Yellen (JME
1989)
– Use both bilateral and multilateral data on similar issue
Rose: Comments on Ghosh, Qureshi and
Tsangarides
2
Praise 1
• Good question, well-motivated
– Divergence between different bilateral US$
regimes a great example
– Notice though: need an anchor for relevance
• Nice encompassing approach
– Reproduce weak multilateral and then get strong
bilateral results
• Easy to replicate (with their data)
Rose: Comments on Ghosh, Qureshi and
Tsangarides
3
Praise 2
• Admirable sensitivity analysis
– Cut data by income, change estimator…
– Current account/trade balance, normalization
(GDP/Trade) issues handled well
• Lithuania natural experiment (2002 switch
from US$ to €)
• Ancillary support (real exchange rate
movements)
Rose: Comments on Ghosh, Qureshi and
Tsangarides
4
What does it Mean?
• Suppose accept premise that relationship
exists in bilateral but not multilateral data
• What does this mean?
– Empirical Options
• Measurement Error: multilateral regime classification
sucks, bilateral better
– Plausible? Bilateral classifications derived from multilateral
• Sample size: too little multilateral data?
– Too much bilateral? (left-handed labor economist)
Rose: Comments on Ghosh, Qureshi and
Tsangarides
5
Smaller Criticisms: 1
• CFA franc zone experiment seems contrived,
not compelling
– France reliably pegged to DM, guilder, … pre-Euro
– Ditto 1999 creation of Euro
• Does BOR data go back to 1980 reliably?
• Current accounts more interesting than trade
imbalances (but highly correlated)
Rose: Comments on Ghosh, Qureshi and
Tsangarides
6
Smaller Criticisms: 2
• “Multilateral” better than “aggregate”
• A good graph here would beat pages of
regression coefficients
Rose: Comments on Ghosh, Qureshi and
Tsangarides
7
Soft Criticism 1: Why use
Regime Classifications at All?
• Instead of using three bins (fix, intermediate,
float), why not use continuous measure of
exchange rate volatility?
– Original motivation is whether more flexibility
affects adjustment speed
Rose: Comments on Ghosh, Qureshi and
Tsangarides
8
Soft Criticism 2: Incomplete
Model of Trade Balance
• Model links trade balance only to exchange
rate regime, a lag and interaction
• Mis-specification orthogonal to regime
interaction?
• Why not include other determinants of
external account (model-dependent: output,
real exchange rate, more lags for RY ’89;
relative wealth, non-tradeables, etc)?
Rose: Comments on Ghosh, Qureshi and
Tsangarides
9
Soft Criticism 3:
Much Ado about Little?
• Many differences are economically small
– Many half-lives are just plain small!
– Ex (pp 14-15): half-life of trade imbalance ≈
•
•
•
•
1.2 years under fix
.9 years under float (plausible?)
So … difference is small (plausible? important?)
Small regime differences also on p21; .1 year
• (But this is necessarily a short-run question)
– All real exchange rates float at low frequencies
Rose: Comments on Ghosh, Qureshi and
Tsangarides
10
Hard Criticism 1: Does the
Effect Work too Well?
Shouldn’t high inflation make nominal
exchange rate regime irrelevant?
Obs
OLS
DJ
CPFE
DJ
CPFE/TE
DJ
OLS
DF
CPFE
DF
CPFE/TE
DF
All
258,075
-.13**
(.01)
-.11**
(.02)
-.11**
(.02)
-.12**
(.01)
-.10**
(.02)
-.10**
(.02)
>10%
25,461
-.12**
(.04)
-.11
(.13)
-.10
(.13)
-.12**
(.04)
-.14
(.13)
-.13
(.13)
>25%
3,899
-.23**
(.07)
-.69**
(.11)
-.62**
(.17)
-.23**
(.07)
-.69**
(.11)
-.62**
(.17)
Inflation
Critical Negative Interaction (γ3) Effect, Table 7
Country-pairs: a) unrestricted; both with b) moderate; or c) high inflation
Rose: Comments on Ghosh, Qureshi and
Tsangarides
11
Hard Criticism 2:
Sensitivity over Time?
Is exact sample period relevant?
Obs
OLS
DJ
CPFE
DJ
CPFE/TE
DJ
OLS
DF
CPFE
DF
CPFE/TE
DF
All
258,075
-.13**
(.01)
-.11**
(.02)
-.11**
(.02)
-.12**
(.01)
-.10**
(.02)
-.10**
(.02)
1980s
50,943
-.13**
(.02)
-.13**
(.05)
-.13**
(.05)
-.11**
(.02)
-.11*
(.05)
-.11*
(.05)
1990s
78,312
-.17**
(.02)
-.09
(.05)
-.09
(.05)
-.16**
(.02)
-.08
(.05)
-.08
(.05)
2000s
128,820
-.10**
(.01)
-.07**
(.03)
-.07**
(.03)
-.10**
(.01)
-.06*
(.03)
-.06*
(.03)
Inflation
Critical Negative Interaction (γ3) Effect, Table 7
Rose: Comments on Ghosh, Qureshi and
Tsangarides
12
Hard Criticism 3: Are
All Observations Equal?
Weighting by GDP eliminates De Jure Result
Smaller Effect on (more important) De Facto
Weighted
OLS
DJ
CPFE
DJ
CPFE/TE
DJ
OLS
DF
CPFE
DF
CPFE/TE
DF
-.13**
(.01)
-.11**
(.02)
-.11**
(.02)
-.12**
(.01)
-.10**
(.02)
-.10**
(.02)
-.02**
(.00)
-.08
(.05)
-.08
(.04)
-.05**
(.00)
-.13**
(.05)
-.12**
(.04)
Critical Negative Interaction (γ3) Effect, Table 7
Regressions: a) unrestricted; b) weighted by real GDP
Rose: Comments on Ghosh, Qureshi and
Tsangarides
13
Hard Criticism 4:
Using Too Much Data?
Restricting to observations with an anchor
Reduces/Eliminates Interaction
With an
Anchor
OLS
DJ
CPFE
DJ
CPFE/TE
DJ
OLS
DF
CPFE
DF
CPFE/TE
DF
-.13**
(.01)
-.11**
(.02)
-.11**
(.02)
-.12**
(.01)
-.10**
(.02)
-.10**
(.02)
-.07**
(.01)
+.00
(.02)
+.00
(.02)
-.06**
(.01)
+.03
(.03)
+.03
(.03)
Critical Negative Interaction (γ3) Effect, Table 7
Regressions: a) unrestricted; b) with one anchor
Rose: Comments on Ghosh, Qureshi and
Tsangarides
14
Basic Problem of Interpretation
• Country can choose a single monetary regime,
but still has many bilateral exchange rates
– US$ does not float freely against RMB
– But US$ floats freely against €
– Policy-induced flexibility is multilateral, not bilateral
• Seems natural to focus on one partner with
whom have most significant/explicit
arrangements
– US floats against €
– China manages RMB against US$ (an anchor)
– (But … why throw away other bilateral information?)
Rose: Comments on Ghosh, Qureshi and
Tsangarides
15
Summary of Critique
1. Smaller
– Why Use Regimes instead of Variability?
– Silly Model of Trade Balance
– Empirically Results are Modest
2. Bigger
– Inflation Results Worrying: too good
– Unimportant observations too important (early years;
GDP-weighting; non-anchor: non-anchor)
3. What does it mean?
– Country has one monetary policy, many bilateral
exchange rates
Rose: Comments on Ghosh, Qureshi and
Tsangarides
16
What Would I do Differently?
1. Present and discuss these problems
2. Argue that they’re not a big deal
Rose: Comments on Ghosh, Qureshi and
Tsangarides
17