Use of Structural Equation Modeling to examine the relationships between Trade, Growth and the Envir

1
Use of Structural Equation Modeling to examine
the relationships between Trade, Growth and
theEnvironment in developing countries.

• Anna Kukla-Gryz
• Department of Economics, Warsaw University, Poland

2
Plan of the presentation

• Trade, Growth the Environment what do we
  know?

• Questions and Problems

• Formulation and Estimation of Structural
  Equation Model

• Conclusions

3
Environmental Kuznets Curve Hypothesis
4
Trade, Growth and Environment (1)
Openness to International Trade
?
Increase in Incomes
EKC Hypothesis
Environmental quality
5
Trade, Growth and Environment (2)
1) Pollution Haven Hypothesis a reduction in
trade barriers leads to a shifting of
pollution-intensive industry from countries with
stringent regulations to countries with weaker
regulations (from developed to developing
countries)

• Race-to-the-Bottom Hypotheis developing
  countries lower their
• environmental standards to attract
  international business

Is - for developing economies - openness to
international trade good or bad for the
environmental quality?
6
Problems
1) Problem with agregation of environmental
indicators (water quality, air quality,
e.t.c.). 2) Differences in the deffinitions
across countries particularly in developing
countries.
7
Advantages of Structural Equation Modeling (SEM)

• Estimation of latent variables (factors), e.g.
  environmental quality, measured
• by many indicators.
• - Estimation of error terms on each observed
  factors indicator. As a result,
• path coeffcients are unbiased by error terms
  which increase the comparability
• of the data between the countries.
• - SEM allows combining many structural
  relationships into one model giving
• a possibility of including many mechanisms in
  one model, e.g. between openness
• and economic growth, openness and
  environmental quality, economic growth
• and environmental quality.

8
Description of the formulated structural equation
model
The structural equation model The structural
equation model specifies the causal relationships
among the variables, describes the causal
effects, and assigns the explained and
unexplained variance.
The measurement model for dependent latent
variable (factors) The measurement model
specifies how latent variables depend upon or are
indicated by the observed variables.

• h is a m x1 random vector of latent dependent
  variables
• x is a n x1 random vector of exogenous
  variables
• y is a p x1 vector of observed indicators of
  the dependent latent variables h
• e is a p x1 vector of measurement errors in y
• Ly is a p xm matrix of coefficients of the
  regression of y on h
• G is a m xn matrix of coefficients of the
  x-variables in the structural relationship
• B is a m xm matrix of coefficients of the
  h-variables in the structural relationship
• is a m x1 vector of equation errors (random
  disturbances) in the structural relationship
• between h and x

9
Models estimation
The goal of estimation is to produce a covariance
matrix s(q) that converges upon the observed
population covariance matrix, s, with the
residual matrix (the difference between s(q) and
s) being minimized The general form of the
minimization function is Q (s - s(q))W(s
- s(q)) s - vector containing the
variances and covariances of the observed
variables s(q) - vector containing corresponding
variances and covariances as predicted by the
model W - weight matrix, W is chosen to
minimize Q The weight matrix, W, corresponds to
the estimation method chosen.
10
Indicators of the latent variables

• Urban population as a percentage of total
  population
• Literacy rate of 15-24 years old
• Non-agricultural workers, percentage of total
  labour force
• Mortality rate in children under 5 year olds
• Health-adjusted life expectancy (HALE)
• Immunization rate for DPT in one-year-olds
• Immunization rate for measles in one-year-old
• Percentage of population with acces to improved
  water source
• Percentage of population with acces to improved
  sanitation
• Fertilizer use intensity
• Total Forest area, average percentage change in
  1990-2000
• Carbon dioxide emissions per capita

structural changes (dev)
health care quality (health)
environmental quality (env)
11
Economic indicators (exogenous variables)
GDP PPP per capita Foreign direct investment
intensity International Aid received by
country Openness Freedom Index Export to
developed countries percentage of total
export Export of manufactured goods (5-8 SITC
Rev. 3, without 68) percentage of total export
120 developing countries, without CEE countries,
year 2000
12
Conceptual path diagram of the model
Goodness of Fit Index (GFI) 0.793 Adjusted
Goodness of Fit Index (AGFI) 0.664
Chi-Square158.26, df117, P-value0.00666,
RMSA0.054
13
Estimation Results (1) Measurement Equations
lit 0.573dev, Errvar. 120.711, R2
0.544 (10.460) (8.320)
agri 1.000dev, Errvar.
144.701, R2 0.752
(5.803) urban 0.744dev,
Errvar. 181.799, R2 0.572
(14.442) (8.853)
structural changes
Hale 14.976health, Errvar. 30.528, R2
0.685 (29.463) (4.094)
dpt 24.545health, Errvar.
142.139, R2 0.557 (8.293)
(7.397) measles
22.607health, Errvar. 134.810, R2 0.529
(8.206) (8.565)
um5 1.000health, Errvar. 0.0619 ,
R2 0.827
(3.314)
health care quality
Fert 29.749env, Errvar. 5312.637 , R2
0.304 (3.980)
(4.510) Water 8.556env,
Errvar. 90.521 , R2 0.680 (4.775)
(5.753) Sanit
9.043env, Errvar. 338.102, R2 0.388
(4.528) (8.367)
Forest 0.431env, Errvar. 3.225 , R2
0.131 (3.117)
(4.370) co2 1.000env, Errvar. 7.146 ,
R2 0.268
(2.184)
environmental quality
14
Estimation Results (1) Structural Equations
health 0.0243dev 0.00572GDPpc
0.00389ex_manu 0.00376oda 0.0202fi
(8.689) (0.200) (3.720)
(2.217) (0.687) Errorvar. 0.00958
(0.431), R2 0.968 dev
6.908GDPpc - 0.0733exdev 67.290fdigdp -
0.0701ex_manu 0.648fi 14.660open
(13.493) (-1.385) (1.611)
(-1.561) (0.696) (3.706) Errorvar.
106.611 (4.796), R2 0.757 env 0.0763dev
- 0.0577GDPpc 0.00651exdev 0.0102ex_manu -
1.078open (4.207) (-0.594)
(1.509) (2.724) (-2.569)
Errorvar. 0.125 (0.748) , R2
0.952
15
Estimation Results (2) Indirect and Total
Effects of Economic Indicators on latent
variables
Indirect Effects GDPpc
exdev fdigdp ex_manu oda
fi health 0.168 -0.002
1.632 -0.002 - -
0.016 (7.449) ( -1.457)
(1.583) (-1.470)
(0.696) env 0.527 -0.006
5.133 -0.005 - -
0.049 (4.135) (-1.384)
(1.566) (-1.479)
(0.688)
Total Effects GDPpc exdev
fdigdp ex_manu oda fi open health
0.173 -0.002
1.632 0.002 0.004 0.036
0.356 (9.971)
(-1.457) (1.583) (2.185)
(2.217) (1.136) (3.691) dev
6.908 -0.073 67.290 -0.070 -
0.646 14.660
(13.493) (-1.385)
(1.611) (-1.561)
(0.696) (3.706) env 0.469
0.001 5.133 0.005 -
0.049 0.040
(5.099) (0.188)
(1.566) (1.449)
(0.688)
(0.105)
16
Estimation Results (3) Total Effects of Economic
Indicators latents idicators
GDPpp exdev
fdigdp ex_manu oda
fi open Fert 13.958
0.027 152.687 0.144 - -
1.470 1.183 Water 4.015
0.008 43.917 0.041 - -
0.423 0.340 Sanit 4.243
0.008 46.412 0.044 - -
0.447 0.360 Hale 2.595 -0.027
24.444 0.033 0.056 0.537
5.325 dpt 4.253
-0.044 40.062 0.054 0.092
0.881 8.728 lit 3.958 -0.042
38.557 -0.040 - - 0.371
8.400 um5 0.173
-0.002 1.632 0.002 0.004 0.036
0.356 measles 3.917 -0.040
36.900 0.050 0.085 0.811
8.039 agri 6.908
-0.073 67.290 -0.070 - -
0.648 14.660 Forest 0.202 0.000
2.212 0.002 - - 0.021
0.017 urban 5.140 -0.055 50.071
-0.052 - - 0.482 10.908
co2 0.469 0.001 5.133
0.005 - - 0.049 0.040
17
Covariance Matrix of Exogenous Variables
GDPpp exdev
fdigdp ex_manu oda fi
open GDPpc 7.738
(9.692) exdev 12.417 384.190
(2.363) (8.662) fdigdp
0.010 0.101 0.001
(1.584) (2.431)
(6.237) ex_manu 35.747 26.539
-0.134 753.320 (5.458)
(0.590) (-2.509) (9.786) oda
-26.778 0.911 0.215
-198.670 596.480 (-5.408)
(0.025) (2.164) (-4.011)
(5.235) fi -1.628 -4.859
-0.009 -1.046 -3.941
1.712 (-4.783) (-1.809)
(-3.166) (-0.316) (-1.630) (10.685)
open -0.084 0.401
0.002 -0.533 3.037 -0.075
0.090 (-1.275) (0.848)
(3.178) (-0.770) (3.660) (-2.081)
(6.282)
18
Openness and GDPpc in developing countries, year
2000
19
GDPpc, carbon dioxide emissions per capita and
forests average percentagechange in 1990-2000,
in developing countries, year 2000
20
Conclusions (1)
The results show that we should be more skeptical
about the existence of a simple and
predictable relationship between openness to
international trade and per capita income.
Not significant effects of both export to
developed countries (ex_dev) and openness on
the quality of the environment do not support
the pollution haven hypothesis.
21
Conclusions (2)
FDI, openess and ex_dev are possitivelly
correlated with each other. These results
suggest that in the analyzed developing
countries, in 2000, FDI went to more open
economies and came from export-oriented foreign
firms. Further, the estimation reslults show that
the only effect of FDI on environmedntal
quality was increased fertilizer use intensity.
Total effects of GDPpc on all environmental
quality indicators are both negative and
positive. Negative, through increase in carbon
dioxide emission per capita and fertilizer use
intensity. Positive, through increase in the
percentage of the population with access to
improved water source and sanitation and through
increase in total forest area.
22
.
Thank you for your attention !