Transcript Is Trade Good or Bad for the Environment? Sorting Out the Causality

• Young’s Theorem:
– For some function F(x1,x2,x3,...,xn) with cross
partial derivatives Fji and Fij that exist and are
continuous, Fij=Fji F
• Roy’s Identity
– if V(p1,p2,I) is an indirect utility function then the
Marshallian demand for good i is Xi(p1,p2,I)=-VPi/VI
Is Trade Good or Bad for the Environment?
Sorting Out the Causality
Jeffrey Frankel and Andrew Rose
Review of Economics and Statistics 2005
+
NBER Working Paper 9201 (2002)
Theoretical Precedents
• Environmental Kuznets Curve literature
– Inverted-u shaped relationship between pollution and income
• Pollution Haven Hypothesis
– strict regs in rich countries shift polluting industry to poor countries
• Factor Endowments Hypothesis
– relatively capital-rich countries export pollution intensive goods
• Porter Hypothesis
– strict environmental regulation promotes trade*
Digression: Porter Hypothesis
Claim: regulation can raise profitability and
promote exports
Possible Explanations
1. There are positive spillovers in innovation.
2. There’s a free lunch out there waiting to be
claimed.
3. There are frictions in the system.
4. California effect
Theoretical Precedents
• Environmental Kuznets Curve literature
– Inverted-u shaped relationship between pollution and income
• Pollution Haven Hypothesis
– strict regs in rich countries shift polluting industry to poor countries
• Factor Endowments Hypothesis
– relatively capital-rich countries export pollution intensive goods
• Porter Hypothesis
– strict environmental regulation promotes trade*
• Politics and Environment
– democracy promotes efficient regulation, openness, and income
growth)
– as F&R write: “ what if free-market trade policies are no more
important to growth than free-market domestic policies, but tend to
be correlated with them? (p.11)
Conclusion
• Income, openness, emissions are all
endogenous
This paper’s goal
• Determine whether trade and income growth
have positive or negative impact on
environmental indicators controlling for
endogeneity of each.
Equation to be estimated
EnvDami   0  1 ln  y / pop 90,i   2 [ln  y / pop 90,i ]2
  X  M  / Y 90,i  3 Polity 90,i   4 ln LandArea / pop 90,i  ei
• EnvDam - one of seven different measures of environmental damage,
• ln(y/pop) – log of 1990 real GDP per capita,
•[X+M]/Y – ratio of nominal X and M to GDP (openness)
• polity – how democratic is the structure of the government
• LandArea/pop – per capita land area
• e – residual representing other causes
Data
• Dependent variables: cross country data for 1990 for
7 different environmental indicators:
– Concentrations in micrograms per cubic meter (averaged
across a country’s measuring stations and cities) of SO2,
PM, NO2;
– industrial CO2 emissions per capita (in metric tons),
– average annual percentage change in deforestation for
1990-1995,
– energy depletion (=unit resource rents x physical
quantities of fossil fuel energy extracted),
– % of rural population with access to clean water 19901996.
Independent Variables
• Direct measures:
– [X+M]/Y
• ratio of nominal exports + imports]/GDP
– natural log of 1990 per capital GDP
• real PPP-adjusted
– Polity IV Project indicators of
autocratic/democratic nature of gov’t
• ranges from -10 to +10 with -10=strongly autocratic,
+10=strongly democratic
– ln(land area per capita)
Indirect Measures—Trade Intensity
• fitted [X+M]/Y where fitted[X+M] are
predicted by a Gravity Model.
• Gravity models regress actual pairwise trade
on
– log of distance,
– log of partner country population,
– log of area,
– and dummy variables for common language,
common land border, and landlocked status.
Source: Frankel
and Romer, 1999,
“Does Trade
Cause Growth”
AER p.384.
– “After estimating the gravity model for a large
data set on pairwise trade, we aggregate the
exponent of the fitted values across bilateral
trading partners to arrive at a prediction of total
trade for a given country. The correlation
between actual trade shares and [the] generated
instrument is .72.”
Indirect Measures—Income
(From their working paper:)
“other controls”
ln Y / pop 90,i   0   X  M / Y 90,i  1 ln  pop i  Zi
  ln Y / pop 70,i  1 I / Y i   2 ni   3 School1i   4 School 2i  ui
Initial per
capita income
Investment
Growth rate of
population
Estimates of human capital
investment based on
primary and secondary
schooling enrollment rates
Source: Frankel & Rose
2002 NBER working paper
Results for NO2, SO2, PM
Results for NO2, SO2, PM
• standard errors are largest for PM (not stat sig),
• then NO2 (moderately stat sig),
• then SO2 (strongly stat sig).
Results for NO2, SO2, PM
• Openness has neg impact on each type of
emissions.
Results for NO2, SO2, PM
•Polity: improved governance has a beneficial effect.
Results for NO2, SO2, PM
• Just in terms of signs, EKC does seem to be present;
but, again, isn’t stat sig for PM, and only
moderately so for NO2.
Results for NO2, SO2, PM
• IV results: similar to OLS results, with diminished
significance levels in some cases.
3A. Results for other environmental measures
• OLS
Beneficial effect of energy depletion and water access
CO2: free-rider problem, global externality
• IV
Detrimental effect of openness on CO2 loses significance
Beneficial effect of energy depletion becomes significant at 10% level
To summarize
• The use of IV to correct for simultaneity can make an
important difference to some results.
• Some evidence that openness reduces air pollution;
• Little evidence that openness causes significant environmental
degradation;
• Exception: carbon dioxide;
• Supportive of the EKC hypothesis;
• Positive effect of democracy on environmental quality.
Extensions: Do some countries have
a “comparative advantage” in pollution?
Version 1---Income
• High income open countries farm out their
polluting production to low income open
countries.
• Test:
– include interaction term:
Openness x Income.
– If the Income version of PHH holds, this
interaction should have a negative fitted
coefficient.
• Sample Results: (SO2 from Table 6, NBER
Working Paper)
Note: sign on
interaction term is
positive for both IV
and OLS!
Similarly, they find
pos. interaction term
for PM, and don’t get
stat sig results on
interaction terms for
the other
environmental
indicators.
• Version 2: Countries endowed with a large supply of
environmental quality become pollution havens,
exporting dirty goods to more densely populated
countries.
• Test by adding
Openness × Land Area/Capita
• Result: coefficients are of mixed signs and are
insignificant
---no evidence supporting the “land area” version of
pollution haven hypothesis.
• Version 3 (Factor Endowments Hypothesis): trade may
lead to an increase in pollution among the capitalendowed countries and a decrease among the laborendowed countries.
• Test by adding (openness × capital/labor)
• Result: coefficients are of mixed signs and are
insignificant
Their conclusions
• “There is no evidence that poor, land-abundant, or
capital-abundant countries use trade to exploit a
“comparative advantage” in pollution.”
4. Conclusions
• This paper models the effect of trade on the environment,
controlling for income and other relevant factors.
• Contribution: address the endogeneity of income and especially
trade (IV drawn from the gravity model)
• Summary of the results:
• Trade appears to have a beneficial effect on some measures of env.
quality;
• Little evidence that trade has a detrimental effect overall;
• Reject the hypothesis of an international race to the bottom driven by
trade;
• No evidence for the pollution haven hypothesis;
• Trade has an indirect effect on the environment through EKC.
• The major example where trade and growth may have the detrimental
effects feared by environmentalists is CO2.
Caveats
• Cross-country vs. panel data
– Unobserved heterogeneity
• Number of observations is small
– As low as 35 for NO2 in IV estimation.
• Test of PHE vs FEH?
– Income and K/L work in opposite direction, and are correlated
• If interact only one at a time with openness, may well find
statistically insignificant results even if both (PHE and FEH) are
acting simultaneously.