Anda di halaman 1dari 61

POLLUTION HAVENS

Or
12/11/2014

FACTOR ENDOWMENT
A STUDY OF AGRICULTURE
SECTOR

SUBMITTED BY: ASFAND YAR TAREEN


REGISTRATION # 28/M.Phil.ENV/PIDE/2012
SUPERVISOR: Dr REHANA SIDDIQUI
A dissertation submitted to the Pakistan institute of
Development Economics, Islamabad in partial
fulfillment of the requirement of the degree of Master of
Philosophy in Environmental Economics

DEPARTMENT OF ENVIRONMENTAL
ECONOMICS

DECLARATION

I Asfand Yar Tareen solemnly declare and affirm on oath that I myself have authored this MPhil
Thesis with my own work and means, and I have not used any further means except those I have
explicitly mentioned in this document. All items copied from internet or other written sources have
been properly mentioned in quotation marks and with a reference to the source of citation.

Asfand Yar Tareen

ii

ACKNOWLEDGEMENTS

Praise is only for Allah who created the Universe and the mankind, it is he who taught what I know
of and what I will know of. After my Allah my achievement till this stage was never possible
without the support of my Parents. Secondly I would like to show my gratitude to my supervisor
Dr Rehana Siddiqui whose concern and engagement made me solve the complicated issues arising
at different stages of my thesis. I also like to show my gratitude to Dr Wasim shahid Malik whose
lectures actually made me understand and use economics. Last but not the least am grateful to all
those who helped me in the process of degree and my thesis.

Asfand Yar Tareen

iii

Contents
List of tables and figures.................................................................................................................v
Abstract .......................................................................................................................................... vi
chapter # 1 ...................................................................................................................................... 1
Introduction.................................................................................................................................... 1
chapter # 2 ...................................................................................................................................... 5
Literature review ............................................................................................................................ 5
chapter # 3 .................................................................................................................................... 11
Theoretical framework................................................................................................................. 11
chapter # 4 .................................................................................................................................... 21
Data & methodology .................................................................................................................... 21
chapter # 5 .................................................................................................................................... 34
Results & discussions ................................................................................................................... 34
Conclusions & recommendations................................................................................................ 48
Appendix ....................................................................................................................................... 49
References .................................................................................................................................... 52

iv

LIST OF TABLES AND FIGURES


Table 4 1 Descriptive Statistics Of Variables ............................................................................................ 25
Table 5 1 Regression Results ..................................................................................................................... 36
Table 5 2 Sensitivity Analysis..................................................................................................................... 43
Figure 4 1 Agriculture share of GDP % 2011 ........................................................................................... 22
Figure 4 2 Agricultural Pollution Concentration of Carbon di oxide Equivalent by Country 2011 ...... 23
Figure 4 3 Distribution of Agricultural Pollution Concentration C02 Equivalent ................................. 24
Figure 4 4 Distribution of Agricultural Pollution Concentration log - C02 Equivalent ......................... 24
Diagnostic Test 1 Time Fixed Effect ......................................................................................................... 49
Diagnostic Test 2 Pesaran's test of cross sectional independence ............................................................ 49
Diagnostic Test 3 Wooldridge test for autocorrelation in panel data ....................................................... 50
Diagnostic Test 4 Modified Wald test for groupwise heteroskedasticity in fixed effect regression model
.................................................................................................................................................................... 51

ABSTRACT

This paper investigates the pattern of trade for agriculture sector using dataset of 20 countries
(developed and developing) from 1980-2011 using decomposition method. This method is helpful
in analyzing how different economic factor effect the agriculture sector while decomposing the
impact on pollution into Scale, Composition and Technique effects. Scale variable used in
analysis is the Agricultural output to its overall agricultural land, Composition comprises of the
Capital to Physical labour ratio, and for Technique Research & Development Expenditures in
Agriculture sector is used as Proxy. This variable is the key variable to finding the pattern of trade.
With high level of Research Expenditures leads to increase in income level hence it will be useful
to determine the pattern of trade difference among the developed and developing countries. The
papers finds that Pollution Havens do determine the pattern of trade but also shows that
investment by corporate firms in developing countries are using clean technologies hence reducing
the level of pollution in these developing countries in agriculture sector. And with Trade intensity
pollution concentration are reduced in the agricultural sector thus is beneficial to the environment.

vi

vii

CHAPTER # 1
INTRODUCTION
1.1 STATEMENT OF THE PROBLEM
Some of us are environmentalists, while others are economists, but both are interested in debate of
international trade its impacts in various areas, and its theories. In 1970s these debates lead to its first
stringent environmental regulations in developed countries of the world. Which were continued in the
international trade agreement North American Free Trade Agreement (NAFTA) and Uruguay Round of
the General Agreement on Tariffs & Trade (GATT) and later with creation of World Trade Organization
(WTO). Such debates of NAFTA and GATT have subsided, but the fundamental issue remains alive of
the trade and environment. Korber (1995); Kuznets (1955) discusses the rising inequality from trade
according to them some economies may benefit from trade while to others it may result in losses and these
losses can occur on the environment; Kuznets (1995) also argues that forces that caused inequality
(difference) includes the trade, labour supply, as well as technology.
Apparently there are two theories that conclude this debate Pollution havens and Factor Endowment
environmentalists agreement is on Pollution havens hypothesis, whereas economists agree with factor
endowment. Pollution havens suggests countries with low environmental regulations or having low
income level developing countries gets dirties with trade, factor endowment on the other hands suggests
with free trade countries get cleaner because more and more industries are transferred to countries that are
capital abundant developed countries having strict environmental regulations. Both of these theories have
its merits and demerits and both state that advantage from trade is determined by countrys factors of
production, competitive advantage, or income. Tobey (1990); Grossman, Gene M Krueger (1995) also
agrees with factor endowment as the sole determinant for trade not by differences in their policies.
There are numerous studies conducted to find the effect of trade intensity on Environment; among them are
Grossman, Gene M Krueger (1995); Taylor & Copeland (1994) who developed a theoretical framework
and decomposed the impact from trade intensity into three effects; the Scale, Composition and Technique
1

using concentration levels for SO2 this method is becoming useful to analyse different effects on pollution,
where different interaction terms

are also generated which also reveals insightful conclusions. The

empirical work using the same method of Scale, Composition and Technique was further complemented
by Taylor, Werner Antweiler, Brian R Copeland (2001).

The studies conducted however are not sector specific, we know economy is segregated into
different sectors where agriculture is the main sector for developing countries. Agriculture being
the most important sector of the economy highly contributes to GDP in developing economies. In
Pakistan share of agriculture is 14% where for United States this share is 1%2. Trade effect on
different sectors can be different because each sector rely on different sets of inputs comparative
advantage, capital abundance level as well as different level of Income. Grossman and Krueger
study on NAFTA was based on these three effects (but the composition was specific to only one
country Mexico), however this research will complement the same theoretical framework but with
a set of 20 countries having both developed and developing countries. And it also determines how
trade intensity will affect the agricultural sector while determining the patterns of trade and its
various effects from interaction terms3. Interaction terms will play an important role because trade
intensity alone cannot determine pattern of trade, changes in only trade intensity will show the
change in Scale, Composition, and Technique domestically. The income4 difference exploited
across developed and developing countries in technique effect will be used to find out the effect

Interaction terms are generated using Trade Intensity variable with relative Variables of Scale, Composition &
Technique; whereas relative variables are generated by dividing each country variable with the Average of that
variable dataset.
2
Share of agriculture sector in GDP for each Country explained in coming chapters calculated using this research
data set
3
Interaction terms are generated using Trade Intensity variable with relative Variables of Scale, Composition &
Technique; whereas relative variables are generated by dividing each country variable with the Average of that
variable dataset.
4
For Technique effect Proxy of Research & Development Expenditures are used, this expenditure increases the
level of income and it varies across countries, such variation is exploited to find the Pattern of Trade among both
Developed and Developing Economies.

of trade intensity on environment using agricultural sector and determine whether Pollution
Havens or Factor Endowment determine the Pattern of trade. The reason income difference in
technique effect is used for analysis is because both theories predict that trade intensity will alter
economys composition which depends on countries comparative advantage and both (Grossman
and Krueger 1995) suggests to allow policy to change with level of income.
1.2 RESEARCH QUESTIONS
There are numerous questions often asked when determining the pattern of Trade, Does increased
growth with induced Trade intensity affect the environment? Does trade follow Pollution Haven
Hypothesis or Factor endowment Hypothesis? Do different policies have different effects on
environment?
1.3 HYPOTHESES.
1. Trade intensity negatively effects the environment.
2. Pattern of trade is determined by Factor Endowment Hypothesis
3. FDI negatively effects (Increase in Pollution) the environment in developing countries.
1.4 RESEARCH GAP/SIGNIFICANCE OF THE STUDY
To contribute to the existing research and fill the gap, This study will describe and determine the
environmental consequences of trade intensity using agriculture sector and dividing it into Scale,
Composition and Technique Effects including relative strengths, magnitudes & the pattern of trade;
Pollution Havens or Factor Endowments. Existing contribution to trade and environment is backed by
Azhar, Khalil, & Ahmed (2007); Vilas-Ghiso & Liverman (2007); Zhang (2012), Azhar, Khalil, & Ahmed
(2007) conducted his study in Pakistan which finds the correlation with two dependent pollutants; air and
water; this study is not sector specific which may have different effects from Trade intensity, whereas VilasGhiso & Liverman (2007) performs his analysis on Mexican agriculture, whereas Zhang (2012) research is

on energy sector. In light of above this research will take part in determining trade intensity impact on
agriculture sector, and may contribute to policy.

1.5 ORGANIZATION OF THE STUDY:


Chapter 1 presented the main idea behind the research as well numerous questions often asked by
researchers about trade and environment, the hypothesis and significance of the Study.
Chapter 2 will complements this area of study by discussing various literatures which methodology
is used in other literature, how and what determines the trade patterns as well as discussions of
various results in literature. Chapter 3 will determine the theoretical framework used under study,
whereas data description, procedures used to gather data, and variables definitions will be
explained in Chapter 4, Chapter 5 will show the results and discussions from analysis, and Chapter
6 will contain a conclusions drawn from findings, discussion and recommendation

CHAPTER # 2
LITERATURE REVIEW
2.1 TRADE ENVIRONMENT DEBATE
There are numerous debates conducted between economists and environmentalists to discuss the effects of
trade intensity on environment, some have agreed to the benefits of trade intensity while others demand
protective measures before liberalization. These discussions and negotiations were conducted in North
American free trade agreement (NAFTA) and the Uruguay round of general agreement of trade and tariffs
(GATT). According to Herman Daly (1993), growth is as if risky, will lead to degradation of
environment. These debates were intensified with the creation of World Trade Organization (WTO)
which resulted in improvement in solving disputes among the nations and provided a better platform for
negotiations. In 1994 Esty (1994) highlights these disputes in his popular writing Greening the World
trade, according to him restriction on trade liberalization is considered a consummate evil, but for
environmentalists ultimate good lies in protection of land, water and air. Under these negotiations it was
agreed upon that limiting trade as a tax for enforcing environmental settlement will be beneficial, on the
other hand traders see it a discouragement to view such initiatives as unfavourable trade barriers.
Beyond this political economy argument lies a trans-boundary environmental spill overs create a risk
of market failure that could undermine the international economic order and compromise the gains from
an open world trading system Bhagwati, J Srinivasan (1996), 167. Baumol & Wallace (1988) Rules to
control externalities is important for active markets making, and making environment important part of
trade policy. In these discussions there were also several other issues highlighted individually by countries;
India who asked for intellectual property rights; for Nigeria certain pesticides were banned; according to
U.S policies these pesticides may be environmentally harmful but for poor developing countries where
deaths from malaria is quite high, trade-off for such country might lead to different results. Such issues
often arise time to time in these discussions, where new countries who follow environmental regulation

often come across with several reasons where Environment is considered secondary than growth rate of
their country.

2.2 DIFFERENT RESEARCH METHODOLOGIES


The effect of trade on environment is studied by numerous researchers. They have used different theories
and methodologies to prove the effect of trade intensity on economy as well as on environment, there
empirical testings show different results based on their analysis and type of country or industry used in
dataset; Tobey (1990) used the HOV Model for its empirical analysis, he uses 11 resource endowments of
different industries that relies heavily on resources from environment such as mining, metals, paper pulps,
chemicals etc ; this HOV model focuses on multiple commodity model &multiple factor of production.
According to him; he found stringent environmental regulations in 1970 have not affected the patterns of
trade such as Pollution havens does not determine the pattern of trade.
Stokey (1998) studies the long run growth using an inverted U-shape relationship between income and per
capita income. He found that tax and voucher scheme seems to have advantage over direct regulation
because of their availability for capital accumulation
Gale & Mendez (1998), discusses the empirical relationship between trade and environment, they included
SO2 in their analysis to capture the effects of scale, trade and policy, they find that increase in country
activity has negative effect on environment, whose relation was not inverted U-shaped, and also finds that
trade policy didnt show a significant effect; its effects were ambiguous. Pollution rose with its abundance
of capital.
Grossman & Krueger (1991); Copeland & Taylor (1995) uses a decomposition method of Scale,
Composition and Technique effect to determine effect of trade intensity, they found that air quality has
deteriorated with trade. Taylor & Copeland (1994); Taylor, Werner Antweiler, Brian R Copeland (2001);
Vilas-Ghiso & Liverman (2007) also decomposition method of Scale, Composition and Technique effect
to determine effect of trade liberalization, according to this theory countries with higher income specializes

in relatively clean factors of production than countries with lower income level which produces dirty goods,
they also found that factor endowment plays a significant role in determining the pattern of trade
Strutt & Anderson (2000) a case study that investigates Indonesia, it uses a GTAP model, GTAP model is
used in projecting the outcome. This model investigates to explore the policy induced effect from economic
.

activity, projecting its level of changes in composition of output and in techniques of production with its
change in environmental indicators. This model uses air as well as water pollution in its analysis, it projects
different pollutions; for air it includes the carbon di oxide, sulphur , Nitrogen, for water it includes BOD,
COD, DS, SS these pollutions are generated from Paddy rice, livestock, food processing, textiles, clothing,
paper products, chemicals, rubber, manufactures and households, and projections were made for 2010 and
2020. To determine policy induced effect, using same pollutants as well as decomposed data of sectors,
projection of reduction of pollution under Uruguay Round table were also made.
2.3 POLLUTION AS DEPENDANT VARIABLE
There are many studies which use Pollution as Dependent variable. Earlier studies includes Akbostanci,
Tun, & Trt-Asik; McGuire (1982); Trade and Environment, Volume 6: A Theoretical Enquiry (1980);
Copeland & Taylor (1995); Copeland & Taylor (2003); Pethig (1976), Pethig uses one primary sector as
an input, whereas Rudiger uses two primary inputs, Macguire uses a sector model with two primary sectors
as an input. Recent work includes the Copeland and Taylor uses it for many good with many factors and
many different pollutants.

2.4 TRADE PATTERN


Pollution havens as detailed by most environmentalist is the sole determinant of pattern of trade,
Costanza et al (1995) finds simple correlation between national income level of country and its policy level.
Other studies determine the pattern of trade and pollution havens, Mani & Wheeler (1998). They argued
that pollution havens is low wage pollution havens, which shows that environmental regulations increase
with income. According to this research there are two causes of increase in industrial pollution regulations.

Firstl demand for better environment increases with its income level, secondly developed government
institutions are better capable of implementing their environmental regulations. They estimated their results
for dirty industries such as Pulp and Paper and Non-Metallic Mineral products, Iron & Steel, Non Ferrous
Metals, Chemicals, and compared it with clean industries such as Non-electrical machinery, Textiles,
Transport equipment and Instruments. They also found that pollution havens do not have major
significance because a major part of increase in dirty sector share is highly income elastic, and with
continuous income growth elasticity has declined. They also found other factors that have also affected
them such as energy price shocks and energy subsidies in developing countries. Thus showing
environmental regulations increases with income and plays a significant role in shifts from dirty to clean
sector.
Taylor & Copeland (1994); Taylor, Werner Antweiler, Brian R Copeland (2001); Azhar, Khalil, & Ahmed
(2007) decomposes their model into Scale, Composition and Technique effects, Werner Antweiler,
Brian R. Copeland uses the panel data for its estimmates on different cities, they also measure elasticity
estimates to determine the pattern of trade. According to them factor endowment plays a significant role
in pattern of trade. Their results rely on several estimates from effect of Trade liberalization, which includes
the Trade induced compostion effect, and results from various interaction terms. In the end they argued that
trade liberlization reduces Pollution intensity and factor endowment plays a significant role in transfer of
industries from poor to rich countries.
2.5 STUDIES ACROSS PAKISTAN
Azhar, Khalil, & Ahmed (2007) conducted their research in Pakistan they use the same method of
decomposing its model into Scale, Composition and Technique effect which uses two pollutants water
and air pollutant, they use Vector error correction method to find the relationship, according to them air and
water pollution holds a significant effect on level of pollution between the selected countries.

Trade intensity or Trade liberalization has been an important variable to determine the effect of Trade in
analysis, several measures have been used to find the effect of trade liberalization Leamer (1988), among
8

them is the most commonly used (Exports + Imports)/Gdp, this variable is used in known studies as given
above this includes Taylor, Werner Antweiler, Brian R Copeland (2001); Azhar, Khalil, & Ahmed (2007).
Whereas Acharyya (2009); Gamper-Rabindran & Jha (2004) uses dummy variable before and after
liberalization period to measure the effect on environment.
2.6 MEASURE OF FDI TO FIND THE EFFECT OF TRADE ON ENVIRONMENT
FDI is one of the greatest source of transfers from one country to other. It can be in the form of FDI flows
i.e. in form of income and FDI stocks that can be in form of machinery equipment etc. FDI does have a
significant effect on countries, if these transfer are in the right clean industry it will results in expansion of
industry and reducing the overall pollution level in the world as well as in the economy and vice versa. FDI
can benefit in various forms it includes capital transfer, skills and technology, market access and
promotion. Studies which use FDI to measure effects on environment are Acharyya (2009); Damijan et al
(2003) Acharrya uses the industry level panel data to find the effect of FDI he find co-integration to estimate
the relationship. Results from this study show positive relationship between inflow of FDI and GDP
Growth. He also points out that without having proper empirical estimates on the relationship between
sectorial decomposition of FDI inflows and sectorial decomposition to environmental damages it is
premature to conclude either way, it is because pollution intensities and emissions differ across
sectors.Acharyya (2009), 11. Whereas Damijan uses (Research and Development Expenditures (RnD) and
FDI as its measure in transition economies. According to him technology is being tranferred through FDI,
but on the other hand other than the FDI is the RnD which acts as a vehicle to growth in clean goods, there
are four ways technology can be transferred from foreign investment this includes the 1. Institution effect
2. Competition effect 3. Foreign knowledge effect 4. Training effect.
Muhammad, Samia, & Talat (2011) investigates a nonlinear relationship between FDI effect on
environment using data of 110 developed and developing countries using emissions from energy the linear
and nonlinear terms are included in the data, results show environmental degradation increase with FDI
and that environmental Kuznets curve is relevant in the data set used. Kuznets curve shows that with
increase in income first environmental degradation increases but as time passes and income further
9

increases this results in reduction of degradation of environment. The theoretical framework which shows
environmental Kuznets curve which is inverted U-shaped is possible has to follow certain conditions with
increase in income. Other studies GFredriksson (1999), chap 5 also prove Kuznets curve which uses both
air and water pollution data for developed and developing countries to measure the industrial pollution in
Economic development, they found that air pollution results were consistent with Kuznets curve but water
pollution gave ambiguous results he used total industrial pollution data to the share of manufacturing
output.
CONCLUSION
Above studies show that in some economies, industries and sectors Pollution Haven is proven while in
other factor endowment determines its pattern of trade. These studies are however lacking the important
sector agriculture. Neglecting this sector will results in numerous consequences and raise cost in future and
trade being one instrument that may result in increase in pollution in environment or help reduce pollution.
The two theories Pollution havens or Factor Endowment can be used together to determine the mix of
trade that may not result in losses to environment because the sole purpose of policy would be optimum
output with less damage to the environment or reduction in damage from either policies or mix of policies.

10

CHAPTER # 3
THEORETICAL FRAMEWORK
The review of literature helped in analysis of the theoretical model which was completely explained by
Taylor, Werner Antweiler, Brian R Copeland (2001). This theoretical framework circulates around some
major Questions: How trade effects the environment? Is factor endowment the major source of pattern of
trade for agriculture sector or Pollution haven Hypothesis is the major impediment force in determining
the pattern of trade? And how endogenous pollution policy or change in income will effect the environment
through trade patterns? Trade patterns are determined from two methods Factor endowment as
economists suggests that capital intensive countries will relocate to more developed countries with openness
to trade. Whereas Pollution Haven Hypothesis states, relative countries with lower income, will become
dirtier with trade with relocation to developing countries.
Gale & Mendez (1998); Taylor & Copeland (1994); Copeland & Taylor (1995); Taylor, Werner Antweiler,
Brian R Copeland (2001) decomposed Pollution level into Scale, Composition and Technique effect.
They found positive relationship by exploiting the panel structure. Trade intensity resulting in increase in
income leads to reduction in pollution level. There do exists a positive effect due to scalar increase resulting
in increase in pollution level and composition effect resulting in positive effect, whereas technique
effect results in reduction of pollution level. The technique outpace the scale and composition effect.
They also pointed out trade induced changes in composition of nations output, but there is less evidence to
believe that intensity effects composition of output in all countries equally this is because composition of
output depends on comparative advantage of the countries and their relative strength of the three effects of
scale, composition and technique.
As discussed pollution consequences of income growth depends on trade induced income change created
from capital accumulation, however differences exists; capital accumulation promotes production of dirty
goods and neutral technological progress do not. These consequences are dependent on sources of growth.

11

3.1a ECONOMIC THEORY


There are N agents in an economy, which produces goods from two industries with two primary factor of
production; capital and physical labour K & L. These two industries produce two output but one industry
which is the capital intensive industry produces dirty good whereas the other industry which is labour
intensive industry produces clean good; dirty industry is the industry X and pollution is generated from this
industry as its by product, whereas clean industry Y, there also exists constant returns to scale in production
technology. We describe the unit cost function as:

(, )
(, ).
Y is a Numeraire and relative price of X is denoted by P

= --------------------------------------(3.1)
Since countries differ in their location and trade barriers domestic prices are not identical to World prices

, represents trade frictions and PW is common relative price of X. > 1 if X is imported & < 1 if X is
exported

3.2a POLLUTION ABATEMENT


Pollution emission is denoted by E, which is produced from emission of X, but firms have access to identical
factor intensities as all other activities in the industry, hence we treat X units as inputs in abatement, if firm
has gross output of firm is units and allocates units to abatement then net output is:

= (1 )
= /
Where represents the intensity of abatement. If Pollution is proportional to output and abatement is
constant returns to its activity level, then Pollution emissions can be written as

= ()-------------------------------------- (3.2)
12

Where () is emission per unit of X output and is decreasing in .To reduce pollution government uses
taxes , the firm creates profits for a firm by producing X are given by revenue less factor payments,
pollution taxes and abatement costs. Using (3.1) and our definition of , we may write profits as

= -----------------------------(3.3)
= (1 ) ()
is the net producer price of gross output. Because there exists constant return to scale, hence
the output of an individual firm is indeterminate, for any level of output, and the first order
condition for the choice of is

= / ()----------------------------------------(3.4)
Where

= ( ), / > 0

Then emission per unit of output is

= ( )---------------------------------------------(3.5)

Where

/ < 0. The production side equilibrium condition are (3.2) and (3.4) and the zero profit

and employment conditions are

= (, )

= +

1 = (, )

= + -------(3.6)

3.3a CONSUMERS PREFERENCES


There are two groups in the society that differ in their preferences the Green who are environmental
friendly and the Browns who care less about the environment in their preferences.

13

=
Each consumer maximizes his utility given its pollution level so the indirect utility function for
the th is written as:

(, , ) =

( )
()

---------------------------(3.7)

= (, ), > 0
G is income per capital (So

() is the income per capita which is obtained through agricultural

sector production), Where () price index and k is increasing and concave. Pollution is a pure public
bad, but green suffers a greater disutility than Browns. We can write this as:

=
()
3.4a GOVERNMENT POLICY
Since there are two sets of preferences for environment, Government uses its policy to maximize the level
of Utility for both groups it is possible through:

[ [ + (1 ) )]-------------------------(3.8)
where

is the weight Government puts on Green individuals, but a more realistic logic tells that

Government attitude varies towards caring for consumer preferences varies across regions of both
developed and developing economies, developing economies care less about the environment.
To maximize the utility based on these different groups across different countries. Utility is maximized,
subject to Private sector behavior, production possibilities, fixed world prices and fixed trade frictions
Private sector behavior is represented by research and development expenditure for agriculture sector giving

14

us maximized revenue as

( , , ) Overall income in private sector is Private sector revenue plus

rebated taxes.

= ( , , ) +
First order condition this yields:

/ ()

[ + (1 ) ]
=0

With fixed world prices we have:

= 1/()[
++ ]

( )
()

Rearranging will yield Samuelson rule:

= [ (, ) + (1 ) (, )]
(, ) = ()/ /
Above is the marginal damage per person where

> 0 , Simplifying allows us to rewrite (3.9) as:

= (, )----------------------------------------- (3.9)
Country Type is

= + (1 )
And effective marginal damage is:

(, )
Pollution policy therefore varies to Government type and economic conditions.

15

3.5a DEMAND AND SUPPLY OF POLLUTION


There are different economic factors that affect the supply and demand for pollution, decomposing these
into scale, composition and technique effects is useful in this regard. As discussed previously private
sector demand for pollution is implicitly defined by (3.2), to rewrite demand in a more convenient way for
our empirical work, Economies scale is defined as an economys scale as the value of national output of
agriculture sector at base year world prices Scale is measured as

= + ---------------------------------(3.11)
Pollution Emissions can be written as:

= = ----------------------------------(3.12)
Where is the share of x, above equation shows a decomposition, pollution depends on pollution intensity
of dirty industry e(), the relative importance of dirty industry in the economy

and the overall scale of

the economy S. In differential equation form,

= ^ + ^ + ^ --------------------------------(3.13)
Where hats denote percentage change, the 1st is Scale, it measure the increase in pollution generated if the
economy were simply Scaled up holding constant the mix of goods produced and production techniques

e()" i.e. if the endowments in a particular economy are increased by 10% and if there exists no change
in output or emission level than we would see a 10% increase in pollution level.

Then is the Composition

and technique effects. Composition effect can be defined as if we hold the Scale and emissions level of
economy constant than economy which uses more of these goods pollute more. And all else constant an
increase in emission intensity will increase pollution.
Quantity index is used as a measure of the level of output as Scale, because of the change in prices which
creates opposing effects of Composition and Technique effects, so it makes it compulsory to divide each
into separate determinants. By using (3.6) we can solve for total output share

as function of

16

capital/labour,
by

= /, net producers price and base year prices, Composition of output is given

= (, ).

And we have composition effect given by:

^ = ^ + ^ -----------------------------(3.14)
Next to differentiate we use 3.1 and 3.4 to find

^ = (^ + ^ ^ )(1 + ) ^ ---------------------(3.15)
= ()
Now using 3.1 and 3.5 we find

^ = , (^ + ^ ^ )------------------------(3.16)

Where the elasticity of emission intensity with respect to p/>0. Then combining (3.13-3.16) we obtain
decomposition of demand for private sector.

^ = ^ + , ^ + [(1 + ), + , ] ^ + [(1 + ), + ,/ ]^

[, + , ] ^ )-----------------------------------------(3.17)

All elasticities are positive. If we draw derived demand in {z, T} space, then (3.17) shows that an increase
in Scale, Capital abundance, or the world prices of pollution demand curve to the right. A reduction in trade
frictions shifts the pollution demand curve to the right for dirty good exporter, but to the left for a dirty
good importer. An increase in pollution tax reduces the demand for pollution through two channels. Firstly
it lowers the demand for pollution by raising abatement cost and by lowering emissions level per unit of X
produced. Secondly higher pollution taxes will lower producers price of X and creates a shift in the
composition of output that lowers X output of any given emission intensity. Supply of pollution depends

17

on government policy who sets the price. So decomposition of pollution supply is from 3.1 and 3.10 and
measured as:

^ = ^ + , ^ + , ^ + , ^ --------------(3.18)
, > 0 , > 0
If (3.18) is drawn is two space than increase in investment level (which is the expenditure in research and
development) and relative prices shifts pollution supply upwards. For example if we give more weight to
Greens than Browns, than policy will become more stringent and supply of pollution will shift upward.
Similarly an increase in real income increases the demand for environmental quality shift the pollution
supply upward as well. An increase in pollution supply makes consumption of market goods more
expensive relative to environmental protection. This may create pure substitution effect. ,

>0

3.6a TRADE INDUCED COMPOSITION EFFECT


PROPOSITION:
Consider two economies that differ in their Trade frictions: (i) if both countries export polluting good, then
pollution is higher in country with lower Trade frictions; (ii) vice versa. This isolates the Trade induced
composition effect, although the sign of the effect differs across countries. For an exporter rises with
free trade and raises relative price of dirty good X. This results in a shift of dirty good exporter demand to
the right and pollution supply upward. Pollution demand shifts out towards X; and emission intensities
increase because abatement is now more costly. This shift in pollution supply lessens the increase in
pollution as pure substitution effect of the goods price increase leads the government to raise the pollution
tax. However demand side effect results in supply side substitution effect and also increases pollution.
Consequently emissions must rise. This increase in emission represents the trade induced composition
effect. Proposition therefore also implies that, if we look at cross countries and hold emissions constant,
one may not find openness per se related in any systematic way to emissions. Proposition is helpful in two
aspects. It captures partial effect of Trade liberalization, and the results from proposition are in fact

18

conditional on Trade patterns but Proposition itself is invisible on determinants of Pattern of Trade. These
issues are treated below. If pollution havens determines the pattern of trade in dirty goods, a positive value
would specify that increase in Composition would increase pollution demand hence shifting the curve
rightward which increase concentration level, whereas negative effect that is reduction in demand for
pollution would shift production possibility curve inward hence reducing pollution level Taylor, Werner
Antweiler, Brian R Copeland (2001), 895.

3.7a TRADE INTENSITY IMPACT


To measure the full impact of trade liberalization, one must account for change in real income, the Scale
and its Composition. Differentiating (3.12) with respect to , holding world prices, country type and factor
endowments constant, to find

= 1 ( ) ( ) 3 ( ) ( ) + 4 ----------------(3.19)

A fall in trade frictions produces a scale effect and a technique effect and the trade induced
composition effect. These three effect determine the environmental consequences for trade. Below figure
1 shows the dirty good exporter to a fall in trade frictions, in bottom panel we depict the pollution
consequences of reduction in trade frictions. Before reduction in trade barriers production is at point

A,

whereas world prices is pw and net Price is pN. Since we have assumed that country is exporter of dirty
good and therefore has consumption at point to

A along the economy budget constraint. Note the values in

world prices of domestic output at A measures the Scale of economy.


In the bottom panel we depict the equilibrium level of pollution both before and after trade frictions, as we
know

= (). Hence production is at A and intensity is at e (), pollution is given by EA . when

restrictions on trade falls, world prices and domestic prices becomes equal to each other and production
level moves to point

C and where producers price is PN. At point C real income is higher and there is a

change in the techniques that is used in production. Likewise emission will fall to

e ( )

and overall
19

pollution falls to

EC.

This methodology divides the movement from

EA to EC into three components.

Holding all else constant first through Scale and the Techniques, trade creates a change in composition of
output by movement from A to B. Corresponding to this movement results in an increase in pollution from

EA to EB. Showing same as the Trade induced composition effect. Scale effect is the movement in the
top panel from Point B to Point C. Resulting in Increase in level of pollution is from EB to ES. Finally
technique effect is the fall in pollution from ES to EC as producers switch to clean technology that produces
lower level of emission.
pw
y

pw

PN
C

yC
B
PNt

XA

XC

E=e(C)x

EC
EC
Composition

Technique

EB

Scale

ES

E=e(A)x
Figure 1 Impact of Trade Liberalization on
Environment
Source : Taylor, Werner Antweiler, Brian R Copeland
(2001), 9

20

CHAPTER # 4
DATA & METHODOLOGY
In the study secondary data is used. Selection of most of the data is from Food and Agriculture
Organization (FAO) database. FAO has an extensive database for agriculture, and fisheries, it not
just includes the production data of agriculture but also many other variables of interest, including
agricultural emissions level. FAO has recently updated its emissions database, which is now
extended to 2011 and projections are also made for year 2050. Other useful variables includes,
inward FDI stock and Research & development expenditure. Agriculture sector is considered an
important sector for most economies to grow, this sector contributes to GDP as well as full fill the
world requirement for food. Developing countries heavily rely on agriculture sector, shown in
figure 4.1 in case of Pakistan agriculture contribution to GDP is 14%, for Bangladesh its 10%
whereas for developed countries this percentage is quite less; for Japan and United States Of
America which is 1%.

4.1 SAMPLE
The data acquired for analysis is balanced data analysis. The data used as the sample for my study comprises
a list of countries which are taken from State Bank of Pakistan, it includes the major cumulative trading
partners of Pakistan although these partners do not specify as the agriculture trading partners, but since we
are using the cumulative country list it might compensate for choosing agriculture trading partners, however
even if one do not agree with this list it still is enough for our analysis to find out the pattern of trade because
to determine the pattern of trade we need a list of developed and developing countries which are already
present in this cumulative list. This list contain both developed and developing countries but I dropped 5
countries because data from those countries were missing. The selected countries include; (Australia,
Bangladesh, China, Germany, India, Indonesia, Italy, Japan, Malaysia, Morocco, Netherlands, Pakistan,
South Africa, Spain, Sri-lanka, Thailand, Turkey, United Kingdom, United States, and Vietnam) The list
21

contains major countries but due to data unavailability, five countries were dropped, which includes Saudi
Arabia, Korea, Oman, UAE and Singapore. The data of selected countries varies from 1980-2011. So there
are twenty Cross sections and time series of thirty two years.

Figure 4 1 Agriculture share of GDP % 2011


UK
0%

US
1%
Vietnam Australia Bangladesh
1%
11%
10%

Turkey
5%
Thainland
7%
Srilanka
7%

Germany
0%
India
10%

Pakistan
14%

Spain
1%
South Africa
0%

Morocco
9%
Netherland
1%

Australia
indonesia
Netherland
Thainland

China
6%

Bangladesh
Italy
Pakistan
Turkey

China
Japan
South Africa
UK

Japan
1%

indonesia
8%
Italy
1%

Malaysia
7%

Germany
Malaysia
Spain
US

India
Morocco
Srilanka
Vietnam

Source: Author

4.2 DEPENDANT VARIABLE


4.2.1 POLLUTANT DESCRIPTION
Anthropogenic activities are considered the main cause of pollution, it affect humans through direct as well
as indirect sources FAO (2001); our dependent variable agricultural pollutant is linked readily with two,
main pollutants which are generated from greenhouse gas emissions (GHG), plants absorb up to only 50%
of nitrogen fertilizers applied to land, such inefficiency which results in non-carbon dioxide gases are the
methane (CH4) and Nitrous Oxide (N2O), that are generated from production of crops, livestock and
management of activities. Carbon dioxide emissions are mainly due from cultivated organic soil5. Figure

http://faostat3.fao.org/faostat-gateway/go/to/download/G1/*/E>

22

4.2 shows that china has the greatest share of agricultural pollutant and the second largest share is that of
India whereas Pakistan has a small share as compared to China and India. Such pollution levels are based
on 2011 estimates. Each observation comprises of giga grams of carbon di oxide equivalent concentrations
of 20 countries from period 1980-2011. And since our data set also comprises of China, India, US, UK as
major countries, the level of concentration from these countries remain high than the other developing
countries. The agricultural pollutant is the total agriculture pollutant in Giga-grams of carbon-di-oxide
equivalent, it comprises of Nitrous oxide (N20) and Methane (CH4), which are generated from management
activities, livestock and crops production. Nitrous oxide is also generated from use of burning of fossil
fuels. Since our pollutant are closely linked to these sectors results may be of interest to public policy.
Throughout this analysis and writing the term agricultural pollutant, pollution intensities, and concentration
will mean the same as pollution level.

Figure 4 2 Agricultural Pollution Concentration of Carbon di oxide Equivalent by Country 2011


Spain
1%

South Africa
1%

Pakistan
6%
Netherland
1%
Morocco
1%
Malaysia
1%
Japan
1%

Thailand
UK
3% Turkey 2% US Vietnam
Srilanka
0%
2%
2%
0%

Australia
8%
Bangladesh
3%

China
32%
India
27%

Italy
1%
Indonesia
6%

Germany
2%

Australia

Bangladesh

China

Germany

India

Indonesia

Italy

Japan

Malaysia

Morocco

Netherland

Pakistan

South Africa

Spain

Srilanka

Thailand

Turkey

UK

US

Vietnam

Source: Author

23

4.2.2 CARBON DI OXIDE EQUIVALENT


Carbon di oxide equivalent abbreviated as (CO2 - eq) is used to compare the emission from different
greenhouse gases according to their global warming potential (GWP). It converts gases to the equivalent
amount of carbon di oxide from its global warming potential6
GGCDE= (Giga grams of Pollutant) * (GWP of gas)

I also checked the distribution


of my dependent variable,
Figure 4.3 shows the original
plot of data which is skewed
towards left. To have better
results and fix this distribution I
took the log of my complete data,
which give normal distribution

Figure 4 3 Distribution of Agricultural Pollution Concentration


C02 Equivalent

as shown in Figure 4.4, some


values are shown at the left
hand bar which I consider are
due to temperature variations
which

are

not

observable

through pollutants data

Figure 4 4 Distribution of Agricultural Pollution Concentration


log - C02 Equivalent

http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/Glossary:CO2_equivalent

24

Table 4 1 Descriptive Statistics Of Variables


Variables

OBS

Mean

Std. Dev.

Min

Max

640

10.5

5.770791

20

1980-2011
(32 Years)

640

1995.5

9.240314

1980

2011

Log of Gg of
CO2
equivalent
Absolute
Value/1000
(Hec)
Constant
2005

640

10.412
87

2.046926

2.7656
9

13.635

FAO

640

15145
20

2539146

928

1.34E+07

UN STATS7

Composition
= Capital to
labor
ratio
K/L

USD Million /
absolute
value in 1000
Constant
2005

640

47.675
46

69.26572

0.8237
2

322.1626

FAO

Technique =
RND

USD Million
Constant
2005

588

2.13E
+08

9.30E+08

0.0200
2

7.18E+09

FAO/ASTI/
OECD
STATS

TI= Exports+
imports/Gdp

Export/impor
ts = 1000
USD, GDP =
1000 USD
Constant
2005
USD Million
Constant
2005

640

0.5433
07

0.399823

0.0799
8

2.07254

WDI

526

13660
8.2

342774.9

0.1657
1

2804588

FAO

640

18.995
55

6.322251

6.8474
6

28.9219

GHCNM v3

640

1.0277
31
16.5

0.433186

GHCNM v2

9.240314

0.1201
9
1

2.77778

Time

Average
Degree
Celsius
Coefficient of
variation
1-32

32

Years

Fertilizers

Tonns

460

41210
77

7178563

3.96e+07

FAO

Panel
ID
Year

logCO2
Scale
=
Value Added
Agri/
Agri
land
Area

Fdi
Stock
/Capital
Temperature
Precipitat~n

Units
1- 20

640

Sources
of
Data
Collection
State Bank
of Pakistan

25

4.3 INDEPENDENT VARIABLE


4.3.1 SCALE EFFECT
The first and foremost independent variable in analysis is the Scale of a country which is output of
agriculture, it is derived from value added of Agriculture (which is in absolute terms), divided by
agricultural land (1000 (Ha)) area of particular country i.e. this data is available completely from UN
Statistics.
It measures the increase in pollution as if economy were simply scaled up holding constant the mix of
goods produced and production technique, example if there exists constant returns to scale & endowments
increases by 10% and if relative prices or emission intensities do not change, then pollution will increase
by 10% as well8 .

4.3.2 COMPOSTION EFFECT


Comparative advantage leads to specialization of countries, comparative advantage stems from changes
in relative size of the economic sectors following a reduction in trade barriers9. The change in the share
of the dirty good in national output, if we hold scale of the economy and emission intensities constant, than
economy devotes more resources to producing the polluting good will pollute more10
For Composition effect ratio of capital to physical labour is used; this is an important variable which
determines the comparative advantage of countries. Countries that are highly capital intensive are dirty
industries relative to labour intensive industries . The data of composition effect is derived by obtaining
capital K (USD million) was divided by total agricultural labour force L (Absolute value in 1000) to
give us the capital to labour ratio (K/L). Relative capital abundance is derived by dividing each countrys
capital abundance by average of countries in the dataset of the given year.11.

Taylor, Werner Antweiler, Brian R Copeland (2001), 882


GFredriksson (1999), 2
10
(Taylor, Werner Antweiler, Brian R. Copeland, 2001, p. 882)
11
Taylor, Werner Antweiler, Brian R Copeland (2001), 892
9

26

4.3.2a CAPITAL STOCK


Capital stock is defined as the Gross Capital Stock it includes The activity of crop or animal husbandry,
this measure includes the assets used in the production process covering land development; livestock,
machinery, equipment and structures for livestock. The gross fixed capital stock is the value, at a point
of time, of assets held by the farmer with each asset valued at as new prices, at the prices for new assets
of the same type, regardless of the age and actual condition of the assets. The gross capital stock database
is the sum of individual physical assets = (land development + livestock + machinery + equipment +
structures for livestock)12. The data used from FAO is the Gross investment, which is used to create capital
stock series, depreciation rate for this purpose is taken at 5%. The depreciation rate of 5% is taken for the
sake of simplicity this rate is different across countries as well as across sectors. The procedure to generate
capital stock is given in different papers but the simplest one I find is explained in Hal (2010)

4.3.3 TECHNIQUE EFFECT


Technique effect can be defined as the change in methods of production followed by trade intensity.
Pollution emissions do not necessarily stay constant, its intensity depends on a number of other components
that includes increase in income due to trade intensity, investment liberalization which may bring newer
technologies, relative price of intermediate inputs; a race to bottom in which trade with foreign countries
may result in setting lower environmental standards due to political pressure or for protection of domestic
industries, however if consumer demand cleaner goods, trade liberalization may reduce pollution level,
instead stimulating race to the top and incentive from lobby groups to pressure government to ease
environmental regulations and if sectorial Composition effect results in a shift to more pollution intensive
sectors, both environmental and industry interest intensifies their efforts to receive favours from
environmental policy makers at higher output level more is at stake, both in terms of profits and

12

http://faostat.fao.org/site/660/DesktopDefault.aspx?PageID=660#ancor

Note: Summary statistics are given in Third Chapter of my Dissertation

27

environmental degradation. GFredriksson (1999), 2. In short holding all else constant, reduction in emission
intensity through any possible means such as research and development, policy options or increase in
income will reduce pollution.
Technique effect to be measured is taken from a proxy variable as Research and development expenditure
(RND, USD million), values of Research and Development Expenditure are in Current 2005 which are
necessary to be corrected because all my data are in Constant 2005. To convert from Current to Constant
2005 GDP Deflator 13 is used. Data of GDP deflator are available in World Development Indicators whose
base year do vary country to country. To rescale the 2010 data to 2005 by first creating an index dividing
each year of the constant 2010 series by its 2005 value of each country leaving 2005 values equal to 1. Then
multiplying each year's index result by the corresponding 2005 current U.S. dollar price value which is give
us Constant 2005 values.
Relative RND is obtained by dividing each countrys RND by the average data of dataset for given year,
where data average as described above are my selected countries. Data of RND is available in FAO
database and for most OECD countries data is available at OECD Stats. For RND some of the data were
missing, which were simple ignored.

4.3.4 TRADE INTENSITY


Trade intensity variable which is the total exports of particular country plus total imports of particular
country divided by particular country GDP, exports and imports of agriculture trade was not chosen for the
reason because I didnt find any particular agriculture trade intensity variable14 although that would specify
for agricultural sector. Trade intensity occur due to fall of prices when both domestic and world prices
become equal this is mainly due to fall of trade barriers such as from tariffs or quotas.. Trade intensity
measure is created in various ways some use dummy variables while other use (Exports + Imports) /GDP,

13

https://datahelpdesk.worldbank.org/knowledgebase/articles/114946-how-can-i-rescale-a-series-to-a-different-base-yea>

14

Agriculture Trade intensity Variable was measured and Evaluated, but not presented in the research, the results
from this measure was giving some very unreliable.

28

this measure is most commonly used, it is also used by Taylor, Werner Antweiler, Brian R Copeland (2001);
Azhar, Khalil, & Ahmed (2007).

4.3.5 POLLUTION HAVENS HYPOTHESIS


Trade encourages reallocation of pollution intensive industries from countries having strict environmental
policies to less stringent ones, when tightening of environmental policies creates strong effect in allocation,
trade flows and when such countries having low income do not indulge themselves in highly abatement
technologies they get dirtier with trade this results in an effect known as Pollution havens. This may occur
because countries may have low level of income or they are facing international competitiveness which
may affect their domestic industries hence domestic country lowers their environmental strictness.
Suppose there is an increase in trade intensity between countries having different environmental regulation.
So Countries that tend to have lower environmental standards develop a comparative advantage in dirty
goods production; as dirty industries from high income countries tend to move to countries having low
environmental regulations Taylor, Werner Antweiler, Brian R Copeland (2001), 877

4.3.6 FACTOR ENDOWMENT HYPOTHESIS


Factor endowment suggest dirty capital intensive countries from low income countries will relocate to
relatively capital abundant high income developed countries with trade; resulting in decrease in level of
pollution Taylor, Werner Antweiler, Brian R Copeland (2001), 877

4.3.7 FOREIGN DIRECT INVESTMENT


Foreign direct investment is investment from abroad in ones economy. This investment can come in
economy through different ways, it can be through multinationals corporations investing in particular
country in a particular sector. Foreign direct investment can come in form of flows or in form of stocks;
foreign direct investment through stocks can results in transfer of technology, this technological transfer
depends on the countrys methods of production thus transfer as FDI stock can be clean or dirty according
to donor country method of production.
FAO now maintains FDI data, this data comprises of FDI flows as well as FDI stocks, FDI inflow of stock
data contains almost all the countries, others countries data are easily available on OECD stats. FDI inflows
29

for stocks are divided by each countrys Capital stock in this research, call this measure FDI intensity15.
This data is also available in current 2005 US$ which was converted to constant 2005 using same
methodology as for converting current RND expenditures to constant RND expenditure. FDI database does
contain some negative as well as missing values which are simply ignored. With all the missing and negative
values we have more than 80% of the data which will be sufficient for analysis and may not create bias in
my analysis. FDI intensity variable will play an important role in determining, whether technology
transfers are of clean goods or dirty goods from various countries.

4.3.8 TEMPERATURE & PRECIPITATION


Temperature and Precipitation data are also used in my analysis. The data from all the stations for 20
countries are not available, although each country meteorological department maintains this data but
considering its cost of acquiring it data from all stations could not be used, therefore an open source data
with complete time series is used for analysis. Data for temperature is taken from GHCNM v316, this data is
available in raw form. Data of temperature from each station is taken and yearly averaged to find countrys
average temperature for that year. (Each station data is placed vertically than averaged for given years).
Temperature does play a significant role in determining whether temperature has any effect on pollution
dispersion or not. The data of average temperature17 is in degree Celsius, and for precipitation level we use
the same complete data this data is also maintained as an open source in raw form in GHCNM v218.
Coefficient of variation measure is used to determine the effects from Precipitation in my analysis.

4.3.9 FERTILIZERS
The consumption of fertilizers data is used for sensitivity analysis, this data is also available from FAO
stats website as described. This data however is available from only 1980-2002. Fertilizers data is used as
for testing further proving that factor endowment do not play role in determining the pattern of Trade.

15

Taylor, Werner Antweiler, Brian R Copeland (2001), para 6

16

http://www.ncdc.noaa.gov/ghcnm/v3.php>

17

Averaging country data without taking into considering the areas where agriculture production takes place is short
coming of this research
18

http://www.ncdc.noaa.gov/ghcnm/v2.php>

30

4.3.10 DUMMY VARIABLES


Dummies for developed and developing countries is also used in the analysis, these dummies are used with
variables as an interaction term to find out the flow of increase and decrease in pollution level after trade
intensity or as transfer of pollutants. This interaction term with RND and FDI will be quite useful in
determining pollution havens.
For data analysis; software such as STATA is used for analysis, since my data is panel based that includes
cross sections as my countries and time series as years. There are two main estimation models for panel,
fixed and random effect models.To find whether model is a fixed or random effect model; we use Hausman
test for identification. STATA includes some very extensive packages for estimation of panel. I also
measured the time fixed affect for our data, and cross sectional dependence test for Analysis. The degree
of confidence was measured at three levels 19.To find whether pollution havens play a role or factor
endowment we used FDI as one of our estimations as an explanatory variable. For sensitivity analysis I
used the data from 1980-2002 for estimation, I added one more explanatory variable; the use of fertilizer in
agriculture sector.

ESTIMATING EQUATIONS
1:
/

= + +

/
= + + ( ) + + +

= + .
+ .
+ . + .

+ .
.

19

(*** Significance at 99.1% level of confidence ,** Significance at 95%,* Significance at 90% level of confidence)
31

2:
/

= + +

= + + ( ) + + + ( ) +

= + .
+ .
+ . + .

+ .
.

3:
/

= + +
/

= + + ( ) + + + + + ( )

+ +
= + . / + . / + .
+ . + . / .

contain the climatic variables such as temperature and precipitation which may affect the production
of crops and may result in increase or decrease in pollution level. Above three model contains the
unobservable terms to account for theses exclusions individual effects for given by:

= + +
s the time specific effect, is a site specific affect and

is an idiosyncratic error20 for

observation stations i in city j in country k in year t .

20

Measurement error

32

EXPLANATION OF EMPIRICAL MODELS


The empirical model is derived from theory as shown in (3.12) and (3.13) which separate scale,
composition and technique effect. In the First model A scale (absolute value/1000 Hectares); represents the
output in the agricultural sector which is derived from value added agricultural output divided by total
agricultural cultivated area. The second is the capital to physical labour ratio which represents composition
effect, third shows the technique effect which is obtained from use of proxy variable from research and
development expenditure. The other variables in model comprises of relative terms which is obtained by
dividing each country values by data average.
In the second model; a modification is made to model A; the square of independent variables are taken,
such squares are taken because impact of research and development expenditure on pollution also depends
on the existing level of capital to labour as well as existing expenditures of research and development, this
amendment also represents factor accumulation.
In the third model, another modification is made by taking square of scale, this is taken because
nonlinearity in the impact of scale could arise from non homotheticities in production and consumption
Feasible Generalized Least Square Feasible Generalized Least Square was chosen after confirming from
Hausman test and several diagnostic tests of my analysis

33

CHAPTER # 5
RESULTS & DISCUSSIONS
My main variables in the model include the Scale, Composition and Technique effect, of the three models
described above. The results show both tests fixed and random effects are significant. To choose among the
models selection is made through Hausman test. Fixed effect model using Least Square Dummy variable
approach (LSDV) which show significant results; the probability statistics are significant at *** 99% level
of confidence for all three models. This proves that fixed effect model is the appropriate model for analysis,
but to be sure random effect model is also to be tested for analysis.
As fixed and random effect models are appropriate models for panel data analysis, both are to be tested for
this purpose. I first tested fixed effect model using Least Square Dummy variable approach (LSDV) which
show significant results; the probability statistics are significant at 99% level of confidence for all three
models. This proves fixed effect model as an appropriate model for analysis, but for surety random effect
model is also to be tested for analysis. To check for random effect I use Breusch and Pagan Langragian
multiplier (LM) test, which is applied after estimations for random effect model we find consistent and
significant results for random. Hence proves that random effect model is also the appropriate model for
analysis, therefore we find that there is evidence of significant difference across countries. Since both of
my fixed as well as random effects results are significant I need to identify the appropriate model for
analysis, hence I apply Hausman test for identification. Hausman test helps choose between fixed and
random effect when both fixed and random effects are significantly present. The Null hypothesis for
Hausman test is model preferred is random effect model rather than fixed effect model, it tests whether
errors are correlated with the regressors, and the null hypothesis is the errors are not correlated with each
other. In estimating the Hausman test. I faced problem in STATA which gave an error SUEST (Seemingly
unrelated estimation). SUEST is not solved for panel hence a solution is given by, Schaffer & Stillman
(2011), which gave results at 99% level of significance for all the models that Fixed Effect is the appropriate
model for analysis. Hence we can use Fixed effect model to find the results. The results of Hausman and
34

LR test are given in Table 5.1. I have also used elasticity estimates for my analysis, elasticity

estimates are calculated through mean elasticities, these estimates are useful to determine the
magnitudes effects from increase in scale, composition and technique effect. To check for other
problems several diagnostic tests are also performed to determine the feasible or appropriate test for
analysis21.From Diagnostic test I have come to result that FGLS is my appropriate model for analysis. To
determine the level of confidence I use three degrees of level of confidence.

*** 99.1 percent level of confidence


** 95 percent level of confidence
* 90 percent level of confidence
5.1.1 SCALE EFFECT
The first and foremost independent variable is the Scale variable. In table 5.1 which show results from
regression of FGLS. In the first model I find a strong negative relationships between Scale of Agriculture
activities (Value added agriculture / Agricultural land of particular Country) to its effect on agricultural
pollutant (i.e. Carbon di oxide equivalent concentrations). These results are consistent with all three model
showing a negative relationship. It means that with increase in Scale of agricultural production level of
pollution will decrease. The results of scale are off-course from my expectations; because according to
theory increase in production will increase the level of pollution in the economy, but as my data set contains
developing countries which are labour intensive which produces clean good22, and developed countries
which are technologically advance; there per hectare production remains high leading to low pollution
concentration than their production level hence leading to overall negative scale effect on pollution. The
effect of agricultural production is quite small on pollution reduction. Elasticity estimates for Scale are
measured for all the three models. The elasticity estimates in my model A shows that with one percent
increase in Scale of agriculture there will be 11.2% reduction in agriculture pollution. But in Model B there

21
22

The results from Diagnostic tests are given in the appendix


For details book Trade and Environment by Brian R. Copeland and M. Scott taylor

35

is a little increase in level of pollution this may be because Composition effect outweigh Scale effect or
due to further increase in Production, so in Model B my elasticity estimates show that with one percent
increase in Scale of agriculture there will be a reduction of 9.8% of agriculture pollution. This reduction is
less than Model A but as we move on to model three these estimates are further changed, the elasticity
estimates on one percent increase in production results is 38.3%. So the overall elasticity estimates vary
between 11.2% - 33.3%, these estimates as well as the level of significance differs in random effect and
along the models. In random effect only the second model shows the significant results and have a totally
different sign compared to that in Fixed effect; FGLS model. The elasticity estimates for random effects
are significant in model B which shows that one percent increase in production there would be about 20%
reduction in level of pollution. The model C of fixed and random effects also contain a square of Scale
which shows factor accumulation and that further increasing the Scale2 will create a positive effect of
increase in Pollution. The elasticity estimates vary between 9.7 to 38%.

Table 5 1 Regression Results


Variables
Test Row

Fixed Effect Model


B

Scale

-7.45e-08***

-6.46e-08***

(6.32E-09)

(8.93E-09)

-0.00288***
(0.00115)

-0.0101***
(0.00289)

Scale

Composition

3.81E-05**

Composition2

Technique

Technique
(Tech)

-2.54E-07***
(4.37E-08)
1.64e-11***
(3.59E-12)
-0.0165***
(0.00276)

(0.0000117)

-4.08e-09***

-6.53e-09***

-6.33e-09***

(3.13E-10)

(2.30E-10)
1.15e-18***
(8.31E-20)

(2.26E-10)
1.12e-18***
(8.19E-20)

2.67E-08

1.33E-07***

9.72E-08

(6.33E-08)

(3.40E-08)

-0.00226
(0.00203)

-0.0158***
(0.00258)

(1.03E-07)
-7.36E-12
(7.28E-12)
0.000531
(0.00492)
6.96E-06

6.02E-05***

-(0.0000127)

Random Effect Model


B

4.23E-05***
(0.0000106)

(0.0000199)

-9.90E-10

-2.96e-09**

-5.16e-09***

(5.93E-10)

(1.09E-09)
5.20e-19*
(2.05E-19)

(1.11E-09)
9.08e-19***
(2.38E-19)

36

FDI
stock/Capital
(FDI)
Trade Intensity
(TI)
TI*rel
Composition
TI*rel
Composition2
TI*relTech
TI*relTech2
TI*RelTech*
RelComposition
Temperature
Precipitation
Ti*Tech*
Developing
Hausman/Chi^2
LM Test/Chi^2

-0.0138*

-0.015*

-0.0217**

-0.0041

-0.0121***

-0.00412

(0.00695)

(0.00787)

(0.00815)

(0.00521)

(0.0026)

(0.00487)

-0.616***

-0.697***

-0.818***

0.499***

-0.946

0.488***

(0.118)

(0.123)

(0.124)

(0.143)

(0.1)

(0.14)

-0.0337***

-0.00974*

0.0412**

-0.0272

-0.0779***

-0.0363

(0.00847)

(0.0125)

(0.0148)

0.000498*

(0.000159)
1.11e-08**
(3.98E-09)
5.57e-19**
(1.96E-19)

(0.000223)
1.23e-08***
(1.51E-09)
-2.97e-18***
(2.67E-19)

(0.000228)
1.17e-08***
(1.50E-09)
-2.88e-18***
(2.63E-19)

(0.000192)
2.38E-09
(3.32E-09)
1.76E-19
(1.57E-19)

(0.00943)
0.000706***
(0.000154)
4.67E-09
(2.41E-09)
-1.33e-18*
(5.75E-19)

(0.0203)

0.00109***

(0.0146)
-9.97E-05

(0.000305)
8.87e-09**
(3.34E-09)
-2.39e-18***
(6.97E-19)

-6.40E-11

-1.67E-11

-1.55E-11

-1.73E-11

9.67E-13

3.34E-12

(4.47E-11)
-0.0200**
(0.00723)
0.328***
(0.0658)

(1.77E-11)
-0.0351***
(0.00888)
0.340***
(0.0676)

(1.75E-11)
-0.0311***
(0.00851)
0.298***
(0.0642)

(3.41E-11)
0.00794
(0.0245)
-0.0273
(0.104)

(1.79E-11)
0.0352**
(0.013)
-0.0693
(0.0528)

(3.31E-11)
-0.0023
(0.0221)
-0.0237
(0.0971)

0.000298***

0.000298***

0.000309***

0.0000139

0.000105***

0.0000132

(0.0000522)
56.12***

(0.0000485)

(0.0000487)
33.75***

(0.0000259)

(0.0000123)
40.19***

(0.0000244)

5078.16***

0.000423*

5453.92***

0.000534

5356.48***

ELASTICITY ESTIMATES
Variables
Scale elasticity

Composition
elasticity

Technique
elasticity

0.1124544**
*
(0.00953)
0.1451595**
*
(0.05774)
0.8688637**
*
(0.06667)

Fixed Effect Model


0.0975109**
- 0.3838488***
*
(0.01349)
(0.06591)

Random Effect Model


0.0402506

0.508007***

- 0.8329255***

0.1137916*

(0.14585)

(0.13902)

(0.10227)

1.388144***

- 1.346751***

-0.2106947

(0.04901)

(0.04798)

(0.12611)

(0.09558)

0.2014471**
*
(0.05127)
0.7948659**
*
(0.13014)
0.6290638**
*
(0.23114)

0.1467267
(0.15612)
0.026747
(0.24799)
1.097017***
(0.0065646)

37

FDI Elasticity

TI*Technique*
Developing
N
_cons
LR Test/ chi2

0.0219621**
*
(0.01108)
0.0525609**
*
(0.00921)
588
11.43***
-0.148
775.3***

0.0239137**
*
(0.01255)
0.0525207**
*
(0.00856)
588
11.80***
-0.192
786.3***

- 0.0346692***

-0.0065362

(0.01299)

(0.0083)

0.0545298***

0.0024573

(0.0086)
588
11.98***
-0.187
942***

(0.00458)
588
10.28***
-0.6
31.93***

0.0192992**
*
(0.00415)
0.0185911**
*
(0.00216)
588
10.48***
-0.396
211.6***

-0.0065646
(0.00777)
0.0023207
(0.0043)
588
10.60***
-0.523
48.03***

Note: The 1st value of the variable shows the coefficient value while the second value represents the standard error
whereas Steric shows the level of Confidence

5.1.2 COMPOSITION EFFECT


In model A with one percent increase in composition in the economy results in 14% reduction in overall
level of pollution. The variable capital to labour ratio which is composition effect shows the countrys
comparative advantage, in model B the elasticity estimates further improve resulting in 51% reduction and
in the last model it shows 83% reduction in pollution level. composition effect is an important variable,
this variable shows the cross country differences; comparative advantage According to theory;
developing countries are labour intensive and since they are labour intensive they are involved in production
of labour intensive clean goods, whereas; developed countries production results in production of capital
intensive dirty goods. My analysis contains both developed and developing countries the overall negative
composition effect on pollution results due to large labour force that is involved in agriculture from
developing countries; which produced clean intensive goods whereas developed economies uses advance
technologies or advance capital as a result their production remains high in proportion to creating pollution
resulting in overall reduction in level of pollution in my data set. Further increasing in level of capital to
labour ratio such as the square of composition effect shows will increase concentration levels further
creates a positive relationship hence leading to increase in pollution level.

38

5.1.3 TRADE INDUCED COMPOSITION EFFECT


Trade induced composition effect is the effect on composition due to trade intensity, trade induced
composition effect depends on comparative advantage of the country as well as on the sign of trade induced
composition effect is different across different countries23. If the country is more polluting than the results
from trade will also results in more pollution but if country has less polluting industries it may result in
decrease in level of pollution. Trade Induced Composition Effect are different across different model, in
model A and B it shows significant and decreasing results in pollution level but in model C it shows
significant results with positive increase in level of pollution. The reason that model C has a positive sign
is because Scale was added in it which shows Scale effect over Technique effect.

5.1.4 TECHNIQUE EFFECT


The results from technique effects are also important for my analysis, since these are the extra expenditures
which are used in reduction or improving techniques in production which in case increases the level of
income across different countries. The results from technique effect or trade induced technique effects
are important to determine the pattern of trade. technique effect show consistent results for all three
models, the elasticity from Model A shows that with one percent increase in technique effect creates 86%
reduction in pollution level whereas with other models this estimate reduces up to 13%. The range of
technique varies between 86% to 138%. But with further increase in Research and development
expenditure will result in increase in pollution levels. This may be because the technique effect showed
a negative sign, this sign is consistent according to theory. Composition induced increase in pollution level

5.1.5 TRADE INDUCED TECHNIQUE EFFECT


In table 5.1 Trade induced technique effect is the main variable that determines the patterns of trade whether
its pollution havens or factor endowment, if factor endowment were the accurate measure for our
trade pattern we would see a strong positive relationship between pollution concentration and trade
induced technique effect in developed countries. Rich countries would have comparative advantage in

23

Effect of trade induced Composition effect is explained in the Chapter 3 of Theoretical Framework.

39

dirty good and we would see a positive shift in pollution demand, where as if pollution havens were the
appropriate measure for our analysis we would see a strong positive relationship for developing countries,
developed countries would have a comparative advantage in clean goods leading to negative effect on
pollution concentrations. I use an interaction dummy for Developing countries which show positive
increase in pollution, the elasticity for trade induced technique effect dummy estimates vary between
0.52 to 0.54%. Hence we come to the conclusion that dirty goods production increase with trade
liberalization in developing countries and there is transfer of dirty goods or factor of production to
developing economies. Taylor, Werner Antweiler, Brian R Copeland (2001), 895.
Considering both capital to labour ratio as well as research and development expenditure variable is
important in determining countrys comparative advantage, a positive value would specify that increase in
Composition would increase pollution demand hence shifting the curve rightward which increase
concentration level, whereas negative effect that is reduction in demand for pollution would shift production
possibility curve inward hence reducing pollution level Taylor, Werner Antweiler, Brian R Copeland
(2001), 895.

5.1.6 TRADE INTENSITY


Next in table 5.1 consider the variables of trade intensity, this variable is a combination of (Exports +
Imports) /GDP. This variable is also used with several other variables as an interaction term to determine
the pollution havens or factor endowment for analysis. The sign and level of significance vary across
models. Trade intensity shows a negative relationship between pollution level and trade.

5.1.7 TEMPERATURE &PRECIPITATION


In the model it also indicates two site specific variables the temperature and precipitation, it is observed
that with rise in temperature results in more and more disbursement or dissemination of pollution whereas
high precipitation concentration results in low washing of concentrations24 .

24

Taylor, Werner Antweiler, Brian R Copeland (2001), 894

40

CONCLUSIONS
My finding shows that increase in Trade intensity does reduce pollution levels, and Pollution havens play
an important role in determining the pattern of trade the results are also similar in Acharyya (2009); Strutt
& Anderson (2000) where they find that pollution havens do play role and dirty industries are transferred
to their developing country, agriculture being an important sector of the economy mostly for developing
countries is being polluted with trade and pollution havens explains this phenomena, but the overall
pollution is reduced due to Trade Intensity

5.2 SENSITIVITY ANALYSIS


Dividing the determinants of pollution in Scale, Composition and Technique is useful to determine the
Pollution havens and factor endowment as done in the previous section, but to further test these results
and to do sensitivity analysis another variable of fertilizers consumption data was added this is useful
because earlier analysis contain only one factor endowment variable capital to labour ratio whereas
fertilizers consumption can also increase the level of agricultural pollution; adding this shows factor
endowment expected increase in endowments should have a positive effect or increase in pollution levels.
This fertilizer data was available from 1980-2002. So I used data only from 1980-2002 for my analysis, for
this purpose I also reduced the Time series of my other variables till 2002 and performed my analysis on
this data, this method of performing or reducing time series is also performed in Taylor, Werner Antweiler,
Brian R Copeland (2001). And I have only used Model C to find out the results for sensitivity analysis,
so purposively I am using Trade intensity Research and Development dummy and FDI stock intensity
Dummy for developing for my sensitivity analysis, these all dummies are not used in a single model but is
used in different models.

5.2.1 SCALE EFFECT


Increase in scale will have a negative effect on pollution concentrations as shown in table. This result is
consistent with our earlier results. The reason for its negative sign is the same as developing countries are
more labour intensive whereas developed countries produce more output than level of concentrations and
due to their advancement in technology their concentrations levels are low as compared to their output. The
41

elasticity estimates also show that with one percent increase in Scale the pollution level declines by 32%.
The other Scale measures that include the FDI country dummy given in table below in Column 2 and
Column 5 shows that with increase in Scale of agriculture production there is a reduction of pollution levels.
The elasticity estimates show that with one percent increase in Scale there is a 31% reduction in pollution
concentrations, the same results are also available for fertilizers data and fertilizers country dummy shows
that with one percent increase there is 20% decrease of pollution level according to my data set.

5.2.2 COMPOSITION EFFECT


Composition effect which determines the countrys comparative advantage is an important variable the
results from Composition effect are indifferent than our analysis than before the signs are negative but
are not significant, but the elasticity estimates show significant results hence showing and proving or earlier
analysis that with increase capital to labour there is a reduction in pollution levels. But when I used FDI
country dummy in second model of my sensitivity analysis it shows concentrations are increased with
increase in Composition. It shows 17% increase as more stocks are introduced in model through fertilizers
as well as FDI stock. Since the FDI is entering as a stock (Stock=Capital) if the stock is dirty industry
pollution increased overall. In the next model I used fertilizers in separate model, the composition effect
from using fertilizers data shows that pollution level is reduced less than the other analysis, so with
increased use of fertilizers results in more agricultural pollution due to use of fertilizers but also results in
reduction of pollution level. Hence technically with its one percent increase results in 16% reduction of
pollution level. So I have come to the conclusion that FDI stock transfer to developing countries in more of
a clean technology because it reduces pollution concentration more than the other models.

5.2.3 TRADE INDUCED COMPOSITION EFFECT


Trade induced composition effect shows increase in pollution levels. This result is consistent with my last
effects of trade induced composition effect which shows the same pattern of increase in pollution.

5.2.4 TECHNIQUE EFFECT


Technique effect results are also same as before, increase in research and development expenditure offset
these effects from Capital to labour and from Scale effect creating a negative effect which reduces the
42

level of pollution it shows that with one percent increase pollution concentrations are reduced from 155%
to 190%

Table 5 2 Sensitivity Analysis


FIXED EFFECT
Model
Model FDI
Technique

RANDOM EFFECT
Model Fertilizers

Model
Technique

2.26E-07

2.78E-07

1.46E-07

-1.22E-08

***
(3.24E-08)
1.34e-11***
(2.74E-12)
-0.00392
(0.00217)
9.81E-06

***
(4.80E-08)
1.32e-11***
(3.77E-12)
0.0185***
(0.00536)

***
(2.53E-08)
6.67e-12**
(2.10E-12)
-0.00374
(0.00205)

(7.09E-08)
-2.55E-12
(4.88E-12)
-0.0206***
(0.00358)

-9.17E-05
***

1.21E-05

(0.0000081)

(0.0000267)

(0.00000853)

(0.000017)

-8.83e-09***

-9.12e-09***

-1.00e-08***

-5.80e-09***

(4.66E-10)

(5.10E-10)

(7.10E-10)

1.70e-18***

1.77e-18***

FDI Stock

(2.04E-19)
-0.0281***
(0.00615)

Trade
Intensity
(TI)

Scale

Scale2
Composition

Composition2

Technique
(Tech)
Technique
(Tech)2

Model FDI

Model
Fertilizers

2.34E-07

4.13E-08

(1.49E-07)
-1.32E-11
(9.11E-12)
-0.0358**
(0.0118)

(9.74E-08)
-4.66E-12
(6.81E-12)
-0.00437
(0.00517)

0.000178**

9.21E-06

(9.50E-10)

(0.0000596)
-6.33e09***
(1.33E-09)

(0.0000236)
-6.59e09***
(1.21E-09)

1.89e-18***

1.16e-18***

1.22e-18***

1.31e-18***

(2.15E-19)
0.00163
(0.00791)

(2.96E-19)
-0.0263***
(0.00712)

(2.29E-19)
-0.0111***
(0.00305)

(3.29E-19)
-0.0258
(0.027)

(3.02E-19)
-0.00419
(0.00437)

-1.488***

-0.842***

-0.940***

-1.604

-0.114

-0.0075

(0.17)

(0.127)

(0.125)

(0.1)

(0.119)

(0.108)

0.0720***

-0.0670**

0.0263*

0.154***

0.120**

0.00725

(0.0142)

(0.0255)

(0.0122)

(0.0138)

(0.0419)

(0.0227)

-0.000189

0.00187***

0.000202

-0.00122***

-0.00223**

0.00028

(0.000159)
2.14e-08***
(6.41E-09)
-4.34e-18**

(0.000541)
2.01e-08**
-6.77E-09
-4.77e-18**

(0.000178)
2.62e-08**
(9.27E-09)
-4.49e-18*

(0.000202)
1.16e-08*
(5.58E-09)
-3.36e-18**

(0.000849)
1.90e-08*
(8.44E-09)
-3.27E-18

(0.000316)
1.46E-08
(7.89E-09)
-3.78e-18*

5.29E-05**

TI*Compositio
n
TI*Compositio
n2
TI*RelTech
TI*RelTech2

43

TI*RelComp*T
ech
Temperature
Precipitation
Fertilizers
TI*Tech*Devel
oping

(1.49E-18)

(1.57E-18)

(2.15E-18)

(1.25E-18)

(1.85E-18)

(1.74E-18)

-9.79E-11

-5.37E-11

-1.57E-10

-1.22E-11

-1.18E-10

-2.75E-11

(1.39E-10)
0.0365***
(0.00647)
0.0253
(0.0512)
0.000000114**
*
(7.94E-09)

(1.46E-10)
0.0602***
(0.012)
0.0172
(0.0602)

(2.00E-10)
0.0310***
(0.00578)
0.0661
(0.053)

(1.11E-10)
0.0192
(0.0132)
-0.066
(0.0588)

(1.68E-10)
-0.0109
(0.0204)
0.00333
(0.0901)

(1.59E-10)
-0.00321
(0.0173)
-0.026
(0.0812)

0.000000107***

0.000000190***

3.92e-08***

3.89e-08**

2.35E-08

(7.56E-09)

(2.20E-08)

(7.41E-09)

(1.45E-08)

(4.26E-08)

-0.000514

0.000335*

(0.000439)

(0.00016)

FDI*Developin
g

-0.0499***

0.0235

(0.011)

(0.0274)

Fertilizers*Dev
eloping

-8.79e-08***

3.95E-09

(2.19E-08)
ELASTICITY

(4.29E-08)
RANDOM EFFECT

FIXED EFFECT
Scale elastcity

Composition
elasticity

Technique
elasticity

FDI elasticity

Fertilizers
TI*Tech*Devel
ping

0.32986***
(0.04723)

0.3177832***
(0.05494)

0.179107**

0.7794779***

0.1621584**

(0.09909)

(0.22606)

(0.08879)

1.552276***

- 1.907084***

- 0.205589***

-0.0177428

0.2680595

0.0584065

(0.03577)

(0.10334)
0.9417957**
*
(0.16383)

(0.1707)
1.508969**
*
(0.4978)
1.323473**
*
(0.27752)

(0.13766)

(0.22402)
1.086992**
*
(0.19933)

-0.0385473

-0.0049496

(0.04038)
0.1653126*
**
(0.06173)

(0.00516)

- 1.654812***

(0.08191)

(0.10658)

(0.1172)

0.0351699***

0.0024438***

- 0.031065***

(0.00769)

(0.01183)

(0.00841)

0.4923529***

0.454006***

0.7848296***

(0.03426)

(0.03217)

(0.09068)

1.018415***
(0.16689)
0.0139322**
*
(0.00382)
0.1692285**
*
(0.03195)

-0.023124

0.0150992**

(0.01976)

(0.00719)

-0.1895892

0.0970424
(0.17547)

44

FDI elasticity
Developing

- 0.052**
(0.01147)

Fertilizers
elasticity
Developing
_cons

10.50***
(0.149)

9.853***
(0.279)

0.0244389
(0.02853)
0.2204388***
(0.0)
0.05484
10.41***
(0.116)

sigma_u
_cons

0.0099079
(0.0)
11.10***
(0.335)
0.875***

10.90***
(0.496)

0.10762
10.77***
(0.425)

1.054***
-0.211***

0.991***
-0.171***

sigma_e
_cons
0.232***
N
432
363
460
432
0.347
0.33
st
Note: The 1 value of the variable shows the coefficient value while the second value represents the standard error
whereas Steric shows the level of Confidence

5.2.5 TRADE INDUCED TECHNIQUE EFFECT


Trade induced technique effect is an important variable that helps us determine the pattern of trade. Trade
induced technique effect itself shows significant results with increase in expenditures in R&D results in
reduction of pollution level. It is also used as an interaction term between country dummy for developing
countries. The trade induced technique effect shows a decline of -2.3% but does not show significant
result hence proving that Pollution Havens determine the Pattern of Trade and because there is an
increment of fertilizers in the model it has resulted and more focus on towards factor endowment rather
than Pollution Havens. To further confirm it, the FDI stock dummy for developing can show us the exact
pattern of trade which is discussed next.

5.2.6 FDI INTENSITY


FDI inward stock of agriculture is also an important variable. It confirms my analysis of trade pattern. This
variable data is also used from 1980-2002, by definition this variable is FDI inflows of stock in a particular
country divided by capital level of that country, which is mostly known as FDI Intensity, since the stock
data is available to us we can find whether it has a positive effect on developing and developed economies
or a negative one, FDI inward stock itself along with country dummy can play a significant role in
45

determining, my results it show that FDI has a negative effect it shows increase in FDI intensity reduces
the level of pollution, the elasticity estimates are between 2% to 3%, for my second model of FDI stock
dummy it shows a significant increase in pollution level of 0.0024438% but for developing countries it
shows reduction in pollution level hence showing that FDI inward stock is clean technology the elasticity
is 0.052%, hence proving from FDI table that technology transfer is clean. The reason is if multinational
corporations have common production methods in both developed and developing countries for
engineering, quality control, or other reasons, then the pollution intensity of their production will be
determined by the income per capita of the source country. As a result, a larger multinational presence in a
poor country may mean it is cleaner, all else equal; however, there is an alternative hypothesis working in
the other direction. If multinationals locate in poor countries because of their lax environmental protection,
then we may instead find a positive relationship between foreign direct investment (FDI) and pollution
Taylor, Werner Antweiler, Brian R Copeland (2001), 898. Hence we come to the conclusions that
multinationals who use common production technique produces along with benefits from research produces
negative effect from FDI Intensity to Pollution concentrations. These results are shown in Table 5.2 Column
(3) and (6).

5.2.7 FERTILIZERS USE


Increase in endowments should have a positive effect on pollution concentrations. To find out whether
this is true we add fertilizer consumption variable that is used in agriculture sector, the data of consumption
of fertilizer is available from FAO25 but the data varies from 1980-2002, so for our analysis we will use
same time series for our analysis. We also use the dummy for developing countries to find out the effect
fertilizers have on the pollution level. The results from my analysis show that fertilizers have a positive
effect on increasing the level of pollution, increase in fertilizers consumption increase pollution
concentration, the elasticity shows that with one percent increase in FDI intensity pollution concentration
increases by 49% to 78% and for developing countries I find that pollution concentration is reduced with

25

http://faostat3.fao.org/faostat-gateway/go/to/download/R/RA/E

46

the use of fertilizers probably due to use to enhanced techniques or capital that is transferred in developing
countries.

5.2.8 TEMPERATURE AND PRECIPITATION


My sensitivity analysis show quite different results than my normal model estimations, since in sensitivity
analysis I am using fertilizers data and FDI intensity for developing country dummy, this shows that with
increase in fertilizers temperature increase in temperature results in an increase in pollution dispersion
whereas for precipitation it also shows positive effect, without considering the seasonal variation of
precipitation it shows that increase in precipitation is not causing pollution level to reduce.

CONCLUSIONS
The results from sensitivity analysis shows that Pollution havens determine the pattern of trade.
Industries that are now entering through FDI stocks in the developing countries are corporation
which are using the same clean technology as they do in developed countries. Therefore, the
overall level of pollution is reduced in the world or at least between these trading partners in my
data set.

47

CONCLUSIONS & RECOMMENDATIONS


My research focuses on investigating the effect of trade intensity on the environment using agriculture
sector while determining the pattern of trade. This sector is an important sector for most of the Developing
economies, and since long discussions on effects have shown that trade liberalization may lead to transfer
of pollution or pollution causing industries to poor developing countries. My study has identified this
pattern of trade and may clear it for agriculture sector. Evidently from my data I have come to the conclusion
that most of my variables shows that with increase in Scale, Composition and Technique results in reduction
of pollution level in my data set, hence it shows that my first Null hypothesis is rejected and trade intensity
does not pollute and reduces the level of pollution. And to prove my second hypothesis which determines
the pattern of trade shows that Pollution Havens determine the pattern of Trade, which according to
definition tells us that with increase in trade intensity results in shift of industry from Developed to poor
developing countries but my further analysis also tells me that with increase in FDI stock to Developing
countries results in transfer of clean technology to the industries which is mainly from Corporations using
same factors of production in developed countries, hence rejecting my third claim of my analysis. This
overall analysis hence finds an ideal means for us. Environmentalists can agree on it that FDI for developing
countries are clean technology from multinationals which is an appropriate measure of production in
developing economies. It will not only help reduce pollution but also improve the overall level of trade.

48

APPENDIX
DIAGNOSTICS TESTS
TIME FIXED EFFECT
To test whether time fixed effects are needed, I use J Lloyd Blackwell (2005), it is a joint test to see if the
dummies for all years are equal to zero. Time fixed effect is used while running fixed effect model for
analysis. The probability values for all three models are given below which shows they are significant and
is needed for analysis
Diagnostic Test 1 Time

r2
Chi^2

Fixed Effect

Time Fixed Effects for Normal Models(Data without fertilizers & Year 1980-2011)
A
B
C
0.5303
0.5224
0.5297
199.86***

185.74***

112.58***

CROSS SECTIONAL DEPENDENCE TEST


Pasaran Cross-Sectional Hoyos & Saradis (2006) dependence test is check whether residuals are
Correlated with each other. This cross-sectional dependence is not an issue for micro panels but for long
panels (such as time= over 20-30 years) it may be a problem; to solve Cross Sectional Dependence
problem Driscoll and Kraay are used. These results are consistent for all three models. The results show
that for all three models there does not exist any Cross sectional dependence among the countries, the
results are also same with the Sensitivity analysis data.
Diagnostic Test 2 Pesaran's

test of cross sectional independence

Cross Sectional Dependence for Original Data


Model A
2.564, Pr = 0.0104
ABV off-diagonal elements = 0.468
Model B
1.852, Pr = 0.0640
ABV of off-diagonal elements =

0.466

Cross sectional Test for sensitivity Data


Ti*Tech~Develping
2.090, Pr = 0.0366
ABV of off-diagonal elements = 0.410
FDI_Develop~g
0.869, Pr = 0.3846
ABV of off-diagonal elements =

0.452
49

Model C
1.531, Pr = 0.1257
ABV of off-diagonal elements =

0.468

Fertilizers_Develop~g
4.512, Pr = 0.0000
ABV of off-diagonal elements =

0.417

SERIAL CORRELATION
Serial Correlation is applied to Macro panels, serial correlation problem is common with longer time
series, this problems causes standard errors to be smaller and also resulting in higher R-square, the Null
hypothesis is there is no correlation. To test for Serial Correlation Drukker (2003) Lagrange-Multiplier
test is used for analysis and the results shows that serial correlation is also present in all three models and
even in my sensitivity analysis data.
Diagnostic Test 3 Wooldridge test for autocorrelation in panel data

Serial Correlation Test With Original Data


Model A

Serial Correlation Test With Sensitivity Data


TI*Tech_Develop~g

H0: no first-order autocorrelation


F(1, 19) = 11.598
Prob > F = 0.0030

H0: no first-order autocorrelation


F( 1, 19) = 20.753
Prob > F = 0.0002

Model B

FDI

H0: no first-order autocorrelation


F( 1, 19) = 13.694
Prob > F = 0.0015

H0: no first-order autocorrelation


F( 1, 19) = 22.196
Prob > F = 0.0002

Model C

FERTILIZERS

H0: no first-order autocorrelation


F( 1, 19) = 13.636
Prob > F = 0.0015

H0: no first-order autocorrelation


F( 1, 19) = 22.910
Prob > F = 0.0001

50

HETEROSKEDASTICITY TEST
Problem of Heteroskedasticity may also arise in the data for that purpose I also tested my data for Hetero
for both Original data and found that Problem of Heteroskedasticity was also present in my data. The results
are presented below in the Table.

Diagnostic Test 4 Modified Wald test for groupwise heteroskedasticity in fixed effect regression
model
Test for Hetero for Original Data
MODEL A
H0: sigma(i)^2 = sigma^2 for all i
chi2 (20) = 5086.97
Prob>chi2 = 0.0000
MODEL B
H0: sigma(i)^2 = sigma^2 for all i
chi2 (20) = 4958.00
Prob>chi2 = 0.0000

Test of Hetero for Sensitivity Data


TI*Tech*Developing

H0: sigma(i)^2 = sigma^2 for all i


chi2 (20) = 5482.08
Prob>chi2 = 0.0000
FDI

Model C

H0: sigma(i)^2 = sigma^2 for all i


chi2 (20) = 19648.09
Prob>chi2 = 0.0000
Fertilizers

H0: sigma(i)^2 = sigma^2 for all i


chi2 (20) = 4801.07
Prob>chi2 = 0.0000

H0: sigma(i)^2 = sigma^2 for all i


chi2 (20) = 45087.07
Prob>chi2 = 0.0000

After running Diagnostic tests and keeping the mind that my Hausman test shows Fixed effect as my
appropriate model. And to solve for all the Problems of Heteroskedasticity, Serial correlation, Time fixed
effect I find that these can be solved with Feasible Generalized Least Square method for Panel J Lloyd
Blackwell (2005); the results are explained in the results section.

51

REFERENCES
Acharyya, Joysri. 2009. Fdi, Growth And The Environment: Evidence From India On Co2 Emission
During The Last Two Decades. Journal of Economic Development 34 (1). Chung-Ang Unviersity,
Department of Economics: 4358.
Akbostanci, Elif, G.Ipek Tun, and Serap Trt-Asik. Pollution Haven Hypothesis and the Role of
Dirty Industries in Turkeys Exports. ERC Working Papers. ERC - Economic Research Center,
Middle East Technical University.
Azhar, U, S Khalil, and MH Ahmed. 2007. Environmental Effects of Trade Liberalisation: A Case
Study of Pakistan. The Pakistan Development Review.
Baumol, W, and E Wallace. 1988. Oates (1988). The Theory of Environmental Policy. Cambridge
University Press. New .
Bhagwati, J Srinivasan, TN. 1996. Trade and the Environment: Does Environmental Diversity Detract
from the Case for Free Trade? Harmonization: Prerequisites for Free Trade.
Copeland, Brian R, and M Scott Taylor. 1995. Trade and Transboundary Pollution. American
Economic Review 85 (4). American Economic Association: 71637.
Copeland, Brian R, and M. Scott Taylor. 2003. Trade, Growth And The Environment. Journal of
Economic Literature 42 (May): 771. doi:42.
Costanza, R., K Arrow, B Bolin, and P Dasgupta. 1995. Economic Growth, Carrying Capacity, and
the Environment. Ecological Economics 15: 8990. doi:15.
Damijan, JP, M Knell, B Majcen, and M Rojec. 2003. The Role of FDI, R&D Accumulation and
Trade in Transferring Technology to Transition Countries: Evidence from Firm Panel Data for Eight
Transition Countries. Economic Systems.
Drukker, David M. 2003. Testing for Serial Correlation in Linear Panel-Data Models. The Stata
Journal, 10.
Esty, DC. 1994. Greening the GATT: Trade, Environment, and the Environment.
FAO. 2001. Global Estimates of Gaseous Emissions of NH3 , NO and N2O from Agricultural Land. East.
G.Fredriksson, Per. 1999. Trade,Global Policy and the Environment. World Bank Discussion Paper
No.402, 214.
Gale, Lewis R., and Jose A. Mendez. 1998. The Empirical Relationship between Trade, Growth and
the Environment. International Review of Economics & Finance 7 (1): 5361. doi:7.
Gamper-Rabindran, S, and S Jha. 2004. Environmental Impact of Indias Trade Liberalization.
Available at SSRN 574161.
52

Grossman, Gene M, and Alan B Krueger. 1991. Environmental Impacts of a North American Free
Trade Agreement. National Bureau of Economic Research Working Paper Series No. 3914: 157.
Grossman, Gene M Krueger, Alan B. 1995. Economic Growth and the Environment. Quarterly
Journal of Economicsuarterly Journal of Economics, 35377.
Hal, Robert E. 2010. Creating a Capital Stock Series. ApEc 3006, 3.
Hoyos, Rafael E De, and Vasilis Saradis. 2006. Testing for Cross-Sectional Dependence in PanelData Models. The Stata Journal, 15.
J. Lloyd Blackwell, III. 2005. Estimation and Testing of FIxed-Eect Panel-Data System. The Stata
Journal. http://www.stata-journal.com/sjpdf.html?articlenum=st0084.
Korber, Achim. 1995. Greening the GATT: Daniel C. Esty, (Institute for International Economics,
Washington, 1994), 319 Pp. European Journal of Political Economy 11 (3). Elsevier: 599601.
Kuznets, Simon. 1955. Economic Growth and Income Inequality. American Economic Review 45: 1
28. doi:45.
Leamer, Edward E. 1988. Measures of Openness. University of Chicago Press.
http://www.nber.org/chapters/c5850.pdf.
Mani, M, and D Wheeler. 1998. In Search of Pollution Havens? Dirty Industry in the World Economy,
1960 to 1995. The Journal of Environment & Development.
McGuire, Martin C. 1982. Regulation, Factor Rewards, and International Trade. Journal of Public
Economics 17 (3): 33554. doi:17.
Muhammad, Shahbaz, Nasreen Samia, and Afza Talat. 2011. Environmental Consequences of
Economic Growth and Foreign Direct Investment: Evidence from Panel Data Analysis.
Pethig, Rudiger. 1976. Pollution, Welfare, and Environmental Policy in the Theory of Comparative
Advantage. Journal of Environmental Economics and Management 2 (3). Elsevier: 16069.
Schaffer, Mark E, and Steven Stillman. 2011. XTOVERID: Stata Module to Calculate Tests of
Overidentifying Restrictions after Xtreg, Xtivreg, xtivreg2, Xthtaylor. Statistical Software
Components. Boston College Department of Economics.
Stokey, NL. 1998. Are There Limits to Growth? International Economic Review 39: 131.
Strutt, A, and K Anderson. 2000. Will Trade Liberalization Harm the Environment? The Case of
Indonesia to 2020. Environmental and Resource Economics.
Taylor, MS, and BA Copeland. 1994. North-South Trade and the Environment. Quarterly Journal of
Economics.
Taylor, Werner Antweiler, Brian R. Copeland, M. Scott. 2001. Is Free Trade Good for the
Environment? The American Economic Review 91 (4): 877908.
53

Tobey, James A. 1990. The Effects of Domestic Environmental Policies on Patterns of World Trade:
An Empirical Test. Kyklos 43 (2): 191209. doi:43.
Trade and Environment, Volume 6: A Theoretical Enquiry. 1980. Elsevier Science.
Vilas-Ghiso, Silvina J., and Diana M. Liverman. 2007. Scale, Technique and Composition Effects in
the Mexican Agricultural Sector: The Influence of NAFTA and the Institutional Environment.
International Environmental Agreements: Politics, Law and Economics 7: 13769. doi:7.
Zhang, Youguo. 2012. Scale, Technique and Composition Effects in Trade-Related Carbon Emissions
in China. Environmental and Resource Economics. doi:51.

54

Anda mungkin juga menyukai