Transcript Chapter 1
Is the effect of international trade on
real income the same for all
countries?
• No, of course not!
• Existing studies on trade and income
use
cross-country regressions or
homogeneous panel data models.
Cross-country heterogeneity and
the trade-income relationship
Dierk Herzer
Chair of Economic Development and
Integration, Johann Wolfgang GoetheUniversity Frankfurt am Main, Germany
Contributions
1. Heterogeneous panel cointegration
techniques
•
to estimate the impact of trade on income for
81 developed and developing countries both
individually and as a whole.
2. Variable-selection approach
•
to systematically search for country-specific
conditions that are important factors in
explaining the cross-country differences in
the effect of trade on income.
Organisation
• The impact of trade on income
• The determinants of the impact of
trade on income
• Conclusions
The impact of trade on income
Model and data
•
•
•
•
ln(Yit) = ai + δit + βln(Tit) + eit
Real (PPP) GDP per worker
Trade relative to GDP at PPP
81 countries over the period 1960 to
2003
Unit root tests
Cointegration tests
Estimates of the long-run relationship
The impact of trade on income
Panel unit root tests
Im, Pesaran, and Shin (2003) (IPS);
Cross-sectionally augmented IPS test
proposed by Pesaran (2007) (CIPS)
Variable
Deterministic terms
IPS statistics
CIPS statistics
Levels
ln(Y)
ln(T)
c, t
c, t
-0.06
-0.69
-2.21
-2.16
First differences
Δln(Y)
Δln(T)
C
C
-10.03**
-11.50**
-2.42**
-2.52**
The impact of trade on income
Panel cointegration tests
Madalla and Wu (1999); Larsson et al. (2001);
Pedroni (1999, 2004);
Holly et al. (2009, forthcoming)
Cointegration rank
Standardized panel trace statistics (Larsson et al., 2001)
Fisher statistics (Madalla and Wu, 1999)
CIPS statistic (Holly et al., 2009, forthcoming)
Panel ADF statistic (Pedroni, 1999, 2004)
Group ADF statistic (Pedroni, 1999, 2004)
r=0
6.58**
280.80**
r=1
-0.75
141.84
-2.22**
-3.62**
-2.91**
The impact of trade on income
The long-run relationship between trade and income
• ln(Yit) = ai + δit + βln(Tit) + eit
• Between-dimension, group-mean panel
DOLS estimator suggested by Pedroni
(2001)
allows for heterogeneous cointegrating
vectors
does not suffer from endogeneity bias
• ln(Yit) = ai + δit + 0.165**ln(Tit); t-value: 6.4
The long-run relationship between trade and income
Country
Argentina
Australia
Austria
Barbados
Belgium
Benin
Brazil
Burkina Faso
Burundi
Cameroon
Canada
Chad
Chile
China
Colombia
Costa Rica
Cote d`Ivoire
Denmark
Dominican Republic
Ecuador
Egypt
ln(T)
-0.1208
0.0973
1.0333**
0.8159*
0.9081**
0.1730**
-0.4599*
0.0153
-0.4250**
-0.3859*
0.2260**
-0.9048*
1.0147**
-0.2223**
-0.4991**
0.4908**
0.2625*
2.1883**
0.1727
-1.0723
-0.1405*
Country
El Salvador
Ethiopia
Finland
France
Gambia
Greece
Guatemala
Guinea
Honduras
Hong Kong
India
Indonesia
Ireland
Israel
Italy
Jamaica
Japan
Jordan
Korea, Republic of
Lesotho
ln(T)
0.1573
-0.0756
0.0865
0.4742**
-0.2517*
1.7661
0.1593
-0.3745**
-0.6165**
-0.2436*
0.2059**
0.9489**
0.6044**
1.5980**
0.3654
0.5903**
1.6716**
0.7750**
0.0294
0.1496**
The long-run relationship between trade and income
Country
Luxembourg
Madagascar
Malawi
Malaysia
Mauritius
Mexico
Morocco
Mozambique
Nepal
Netherlands
New Zealand
Nicaragua
Niger
Nigeria
Norway
Pakistan
Panama
Paraguay
Peru
Philippines
Developed countries
ln(T)
0.5847*
0.1402
0.5660**
-0.3663**
0.3932*
-0.4567**
-0.8603**
0.5616**
0.0733
1.7712**
0.1406
-0.0869
0.3121*
0.0194
0.3065**
0.3843
-1.0165**
-0.9748**
-0.3757**
-0.9033
0.7677**
Country
Portugal
Romania
Senegal
Singapore
South Africa
Spain
Sri Lanka
Sweden
Switzerland
Tanzania
Thailand
Togo
Trinidad &Tobago
Uganda
United Kingdom
United States
Uruguay
Venezuela
Zambia
Zimbabwe
Developing countries
ln(T)
-0.16482
-0.0675
0.5676*
0.31041
-0.2050*
1.1401**
-0.1101**
0.2816**
0.5786*
-0.63867**
0.0537
-0.6612**
-0.0887
0.0949
0.4604**
0.2224*
1.0172**
-0.1718
-0.2919**
-0.3792**
-0.0507*
The determinants of the impact of
trade on income
• Bormann et al. (2006); Freund and Bolaky (2008)
the effect of trade on income is negatively related to the
level of regulation.
no robust association between the income effect of trade
and institutional quality.
They use cross-country income regressions that include
interaction terms between trade and a small number of
potential determinants of the income effect of trade.
• In this study, we follow a different approach:
We use a regression model with the estimated income
effect as dependent variable to consider a large number
of possible determinants of the trade-income
relationship.
The determinants of the impact of
trade on income
Variables and data
Variables
ln(gdp)
ln(schooling)
ln(credit)
ln(primaryexports)
ln(firing)
ln(businessfreedom)
ln(monfreedom)
ln(railway)
ln(telephone)
Definition
Log of real per capita GDP (in constant 2000 US dollars at PPP).
Data averaged over the period 1975 to 2003.
Log of the secondary school enrolment rate. Data averaged over the
period 1991 to 2003.
Log of the private sector bank loans/GDP ratio. Data averaged over
the period 1960 to 2003.
Log of the primary exports/GDP ratio. (Agricultural raw materials
exports + food exports + fuel exports + ores and metals exports
divided by GDP). Data averaged over the period 1962 to 2003.
Log of flexibility of firing. Data are from 2003.
Source
WDI 2007
WDI 2007
WDI 2007
WDI 2007
Doing Business,
World Bank (2004)
Log of business freedom. Data averaged over the period 1995 to Heritage Foundation
2003.
Log of monetary freedom. Data averaged over the period 1995 to
Heritage Foundation
2003.
Log of kilometers of railways per square kilometre of land area. Data WDI 2007
averaged over the period 1975 to 2003.
Log of telephone mainlines per 1000 people. Data averaged over the WDI 2007
period 1975 to 2003.
The determinants of the impact of
trade on income
Variables and data
Variables
ln(propertyrights)
ln(corruption)
ln(govstab)
Definition
Log of property rights. Data averaged over the period 1995 to 2003.
Log of corruption. Data averaged over the period 1984 to 2003.
Log of government stability. Data averaged over the period 1984 to
2003.
ln(bureaucratic)
Log of bureaucratic quality. Data averaged over the period 1984 to
2003.
ln(invest)
Log of investment profile. Data averaged over the period 1984 to
2003.
ln(socio)
Log of socioeconomic conditions. Data averaged over the period
1984 to 2003.
ln(democratic)
Log of democratic accountability. Data averaged over the period
1984 to 2003.
ln(intconflict)
Log of internal conflict. Data averaged over the period 1984 to 2003.
ln(extconflict)
Log of external conflict. Data averaged over the period 1984 to 2003.
Dependent variable: βi Impact of trade on income, individual DOLS estimates of the
coefficient on ln(T) over the period 1960 to 2003.
Source
Heritage Foundation
PRS Group
PRS Group
PRS Group
PRS Group
PRS Group
PRS Group
PRS Group
PRS Group
The determinants of the impact of
trade on income
Empirical analysis
• General-to-specific model selection approach
suggested by Hoover and Perez (2004).
• General model
• Eq.1
Eq.2
Eq.3
Eq.4
Eq.5
• Stepwise elimination of insignificant variables
• Eq.1’
Eq.2’
Eq.3’
Eq.4’
Eq.5’
• Final model
The determinants of the impact of
trade on income
Empirical analysis
Independent variable
Dependent variable: βi
ln(primaryexports)
ln(firing)
ln(propertyrights)
Diagnostic tests
Adj. R2
JB (χ2(2))
RESET (χ2(1))
HET
STABILITY
REST
-0.160* (-2.513)
-0.290* (-2.583)
0.485* (2.391)
0.53
2.06 [0.357]
0.11 [0.744]
F(6, 57) = 0.32 [0.923]
F(15, 47) = 1.10 [0.756]
F(15, 42) = 0.60 [0.859]
The determinants of the impact of
trade on income
Empirical analysis
Regressor
ln(gdp)
ln(schooling)
ln(credit)
ln(businessfreedom)
ln(monfreedom)
ln(railway)
ln(telephone)
ln(corruption)
ln(govstab)
ln(bureaucratic)
ln(invest)
ln(socio)
ln(democractic)
ln(intconflict)
ln(extconflict)
Correlation coefficients
ln(primaryexports)
ln(propertyrights)
-0.42** (-3.67)
-0.10 (-0.78)
-0.12 (-0.96)
-0.20 (-1.60)
-0.15 (-1.20)
-0.18 (-1.43)
-0.19 (-1.50)
-0.22 (-1.77)
-0.09 (-0.74)
-0.22 (-1.79)
-0.20 (-1.60)
-0.19 (-1.50)
-0.22 (-1.77)
-0.06 (-0.47)
-0.01 (-0.09)
0.73** (8.47)
0.61** (6.13)
0.27* (2.20)
0.78** (9.68)
0.60** (5.91)
0.52** (4.85)
0.74** (8.56)
0.64** (6.62)
0.58** (5.59)
0.73** (8.36)
0.81** (10.91)
0.82** (11.08)
0.74** (8.57)
0.54** (5.02)
0.51** (4.67)
ln(firing)
-0.45** (-3.93)
-0.33** (-2.79)
-0.28* (-2.34)
-0.42** (-3.65)
-0.29* (-2.38)
-0.42** (-3.68)
-0.45** (-3.95)
-0.39** (3.32)
-0.44** (-3.85)
-0.45** (-3.94)
-0.51** (-4.64)
-0.53** (4.95)
-0.44** (-3.90)
-0.37** (-3.11)
-0.26* (-2.08)
Conclusions
•
A one percent increase in the trade share of
GDP yields, on average, a statistically
significant increase in income per worker of
about 0.16 percent.
• There are large cross-country differences in
the income effect of trade, in particular
between developed and developing countries.
• The cross-country differences in the income
effect of trade can be explained by crosscountry differences in primary export
dependence, labour market regulation, and
property rights protection.
Thank you
for
your attention!