Manufacturing - Finance Comparative Advantage and Global Imbalances by Rui Mao and Yang Yao
ABFER Inaugural Conference discussion by Martin Berka Victoria University of Wellington and CAMA
National University of Singapore Business School May 21, 2013
Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
1 / 16
Comments: Model Do you think it is necessary to have a static model which predicts CA = 0 in a paper on global imbalances? I
Merge the two models into a single one that has CA predictions
I don’t understand some aspects of the modeling setup You seem to have an externality in the model: households choose to give capital, but for 0 direct return I I
I I
I
The return comes in the form of a higher wage next year ”wage return” = 1 + r to facilitate the existence of two assets in a risk-free world This is not a standard competitive equilibrium setup with price-takers It seems you may be solving some kind of central planner’s problem, which in your case doesn’t equal the competitive allocation. In a competitive allocation, households would choose K = 0
Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
2 / 16
Comments: Model Is this a well-defined steady state? I
The within-generation problem has consumption smoothing and growing |CA| F
I
Consumption is constant in each generation: C = 1+1 β Y1 + 1Y+2r
But with a constant interest rate and no risk, your model seems to have a unit root in NFA (NFA → ±∞), possibly as a fraction of ouput
Schmitt-Grohe & Uribe 2003 JIE show that SOE models with incomplete asset markets exhibit dependance on initial conditions, and so are inconsistent with a steady state growth path Stability which prevents transient shocks from having permanent consequences needs to be induced by, e.g., debt-elastic interest rate, convex adjustment costs, etc. What are the stability properties of your model?
Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
3 / 16
Comments: Model
The point you are making with the model is very intuitive and simple In a two-country world with perfect specialization, one country will end up making manufacturing stuff, the other will produce financial services Do you really need a very complex model to say this and bring idea to the data?
Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
4 / 16
Comments: Empirics
Empirical exercise very similar to Chinn and Prasad (2003, JIE) One new variable: Relative labour productivity They cover 89 countries, you cover 24: probably can increase sample? Results are mostly consistent with Chinn and Prasad (2003)
Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
5 / 16
Comments: Empirics
Why this country selection? Two-country model, but data only for small open economies. I
The main feedback loop in your model – the dependance of interest rate on country characteristics (productivity) – need not be satisfied in the data.
There are no CA-creditor countries in the sample except for Korea and Germany (Canada in a some years) Would be nice to see more Asian economies and the US, especially since the usual paradigm in explaining global CA imbalances is ”East vs. West”
Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
6 / 16
Comments: Empirics
Why this country selection? Another reason why it would be nice to see more Asian countries: I I I
I I
Singapore, Hong Kong: world financial centers Also very important manufacturing centers But my guess is that their comparative advantage is in finance, not manufacturing Singapore and Hong Kong run large CA surpluses May not support the theory
Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
7 / 16
Comments: Empirics
Link between Fiscal and CA balances I
I I
”Twin deficit” literature concludes no long-term relationship (some decades +, some −) This is because both CA and FB driven by shocks Persistence and the degree of commonality across countries matter for results
Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
8 / 16
Comments: Empirics
Output per worker is an imprecise measure of what you mean by ”productivity” I
I
Imprecise measure of TFP: Both Y and L respond endogenously to TFP On the flipside, Y /L can also move when TFP doesn’t: F F F
I
non TFP induced changes in K policy changes tax change
These are particularly relevant issues when looking at long (growth) horizons
Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
9 / 16
Comments: Empirics
I use a panel of constructed sectoral TFP levels from another project TFPMan /TFPFin levels do not suffer from the above issues Should give clearer evidence of link between TFP and CA No other variables, but that may work against me Smaller sample: only half of your countries, plus UK, Belgium and Slovenia
Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
10 / 16
TN2 BE .08
.06
GER
SPA
FRA
.00
-.04
-.02
-.02
-.08
-.04
-.04
-.12
-.06
-.06
-.16
-.08
-.08
-.20
.04
.02
.00
-.10
-.10
-.24
-.12
95 96 97 98 99 00 01 02 03 04 05 06 07
95 96 97 98 99 00 01 02 03 04 05 06 07
95 96 97 98 99 00 01 02 03 04 05 06 07
IRE
ITA
NET
.32
.20
.28
.15
.24
.10
.20
.05
.16
.00
.00
95 96 97 98 99 00 01 02 03 04 05 06 07
AUS .04 .00
-.05
-.04 -.10
.12
-.08 -.15
-.05
-.12
-.20
-.16
95 96 97 98 99 00 01 02 03 04 05 06 07
95 96 97 98 99 00 01 02 03 04 05 06 07
95 96 97 98 99 00 01 02 03 04 05 06 07
FIN
SWE
DEN
.16
.20
-.10
.15
.12
95 96 97 98 99 00 01 02 03 04 05 06 07
UK .10
-.15
.05
.10 .08
-.20 .05
.00
.04
-.25 .00
.00 -.04
-.05
-.30
-.05 -.10
-.35
-.10
95 96 97 98 99 00 01 02 03 04 05 06 07
95 96 97 98 99 00 01 02 03 04 05 06 07
95 96 97 98 99 00 01 02 03 04 05 06 07
CZE
HUN
SVN
.16 .12
.18
.20
.16
.15
.14
.10
.12
.05
95 96 97 98 99 00 01 02 03 04 05 06 07
.08 .04 .00 -.04
.10
.00
.08
-.05
.06 95 96 97 98 99 00 01 02 03 04 05 06 07
Martin Berka (VUW)
-.10 95 96 97 98 99 00 01 02 03 04 05 06 07
95 96 97 98 99 00 01 02 03 04 05 06 07
Macro-Fin Comp. Adv and Current Account
May 2013
11 / 16
TFPM /TFPF vs CA: no clear link .12
.08
CA2
.04
.00
-.04
-.08
-.12 -.4
-.3
-.2
-.1
.0
.1
.2
.3
TN2
Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
12 / 16
BE
GER
.08
SPA
.08 .06
.06
FRA
.00
.03
-.02
.02
-.04
.01
.02
CA2
CA2
CA2
CA2
.04 .04
-.06
.00
-.08
-.01
.00 .02 -.02 .00
-.04 .02
.04
.06
.08
-.10 -.10
-.08
-.06
-.02
.00
-.16
-.12
-.08
-.04
-.12
-.10
-.08
-.06
TN2
IRE
ITA
NET
AUS
.01
.06
.00
-.04
.24
.28
.32
.00 -.05
.00
.05
TN2
.10
.15
.20
-.15
-.10
SWE
.08
CA2
.10
.06
-.05
.00
-.12
DEN
.06
-.08
-.04
TN2
UK
.04
-.005
.03
-.010
.02
-.015
.01 .00
.04
-.16
TN2
CA2
FIN
.08
.04
-.04 -.20
TN2
.10
.00
-.02
.02
-.02 .20
.00
.04
-.01
-.06
-.02
.02
CA2
.00
-.04
.04
CA2
.08
CA2
.10
.02
CA2
.03
.02
.16
-.20
TN2
.04
.12
-.02 -.24
TN2
-.02
CA2
-.04
TN2
CA2
.00
-.020 -.025
.04 -.01
.02
.02 -.04
.00
.04
.08
.12
.16
-.030
-.02 -.10
-.05
.00
.05
.10
.15
.20
-.035 -.35
-.30
-.25
-.20
TN2
TN2
TN2
CZE
HUN
SVN
.00
-.04
-.15
-.10
-.10
-.05
.00
.05
.10
TN2
.01
-.05
.00
-.06
-.01
-.04
CA2
CA2
CA2
-.02
-.07 -.08
-.02 -.03
-.06 -.09 -.08
-.04
-.10 -.04
.00
.04
.08
TN2
Martin Berka (VUW)
.12
.16
-.05 .06
.08
.10
.12 TN2
.14
.16
.18
-.10
-.05
.00
.05
.10
.15
.20
TN2
Macro-Fin Comp. Adv and Current Account
May 2013
13 / 16
TFPM /TFPF vs CA in cross-section Cross sections
TN2 Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
14 / 16
Pool vs Fixed effects panel
Dependent Variable: CA (constant not reported) Sample: 1995 2007 Total panel (balanced) observations: 195 Dependent Variable: CA (constant not reported) Period weights (PCSE) standard errors & covariance (d.f. corrected)
Sample: 1995 2007 Total panel (balanced) observations: 195 Variable Coefficient Std. Error t-Statistic Period weights (PCSE) standard errors & covariance (d.f. corrected) TFP_M / TFP_F
Variable
-0.035
0.024
-1.459
Coefficient
Std. Error
t-Statistic
R-squared TFP_M / TFP_F Adjusted R-squared F-statistic
0.011923 -0.035 0.006803 2.328872
0.024
-1.459
R-squared
0.011923 0.006803 2.328872
Adjusted R-squared F-statistic
Prob. 0.146
Prob. 0.146
Dependent Variable: CA (constant not reported) Method: Panel EGLS (Cross-section random effects) Total panel (balanced) observations: 195 Swamy and Arora estimator of component variances Dependent Variable: CAstandard (constanterrors not reported) Period weights (PCSE) & covariance (d.f. corrected)
Method: Panel EGLS (Cross-section random effects) Total panel (balanced) observations: 195 Variable Coefficient Std. Error t-Statistic Swamy and Arora estimator of component variances Period weights (PCSE) standard errors (d.f. corrected) TFP_M / TFP_F 0.078 & covariance 0.030 2.567 Weighted Statistics Coefficient Std. Error
Variable R-squared TFP_M / TFP_F Adjusted R-squared
0.040284 0.078 0.035311
0.030
t-Statistic 2.567
Prob. 0.011
Prob. 0.011
Weighted Statistics
Unweighted Statistics
R-squared
R-squared Adjusted R-squared Sum squared resid
Martin Berka (VUW)
0.040284
-0.108925 0.035311 0.363481
Unweighted Statistics Macro-Fin Comp. Adv and Current Account
May 2013
15 / 16
Comments: Empirics
Results go in opposite direction from yours: pool insignificant, CFX significantly positive Time-series drive the results But with only 15 years of data, your growth story should really come through in cross sectional results This does happen when using Y /L but not when using TFP
Martin Berka (VUW)
Macro-Fin Comp. Adv and Current Account
May 2013
16 / 16
