3. RESEARCH DESIGN AND METHODOLOGY
3.8 Econometric Models
3.8.3 Instrumental Variable Model (IVM)
This study used Instrumental variable model (IVM) to identify and determine endogenous (economic and environmental factors) and exogenous (social factors) effect on resource consumption growth and green environment tradeoffs. Economic and environmental indicators, which are assessed in previous studies such as ESCAP (2011) and WBCSD (2009), were endogenously determined factors for environment protection and sustainability. However, social aspects were excluded in these findings and could not have interlinked by using econometric models. Nevertheless, this study showed social aspects as exogenously determined factors and influenced tradeoffs between consumption growth and green environment problems. Social, economic and environmental indicators were used as guide line to build indicators on factory’s resource consumption intensity. The various indicators integratedinto the socio- eco efficiency were used as the data survey instrument and weighing factors. Correlation levels or strength of relationships between indicators such as the level of green environment and socio- eco efficiency were assessed, characterize and quantified.
In doing so, this study listed indicators and let respondents to determine how each indicator criterion weight during consumption and recycling. Based on indicator criterion, the selection grids have had a scoring system for ranking indicators. The weighted voting can be used numerical systems from 0-10. Where, 1
presents not at all, 5 presents moderate or average and 10 represents maximum or very high. That is the larger number was represented a desirable rating. In some cases, large number may mean "less", for example cost of water or waste removals. In order to set scoring, the researcher sampled to score each indicator against the criteria. Every stakeholder completed how well the indicators would satisfy each criterion. The average score from each respondent is taken. Finally, total and average score is computed based on respondent’s selection and scoring. Accordingly, economical (monthly income), social (culture, religious, gender and etc) and environmental indicator (water consumption quantity and recycles) were weighted highest and recruited as major factors.
However, other factors such as perception, behaviours, awareness, sensitivity and emotionality, ability and willingness, and quantity of resource consumption were included and rated high and found effects on gaps between consumption and green environment problems. To determine indicators, econometric model such as multiple logistic regression models, instrumental variable (IV) and Two Stage Least Square model (2TLS) using Maddala and Guajarati (1983 & 2004) guidelines. Variables consistency, errors, and biasity were checked and tested using maximum likelihood estimation techniques. This study began from assumption and conceptual model, which would capture the interactions of firms, people and environment in different aspects. These aspects included the social, economic and environmental indicators. However, in previous studies, the social aspects were not incorporated into an eco-efficiency. Thus, this study integrated the social aspects to economic and environmental indicators and formulated asocio-eco efficiency framework using an instrumental variable model.
In doing so, it was assumed that there is a relation and interaction between social aspects (consumption culture) and eco- efficiency indicator (economic and environmental) to the green environment. Suppose that social aspects present (Si) and eco efficiency indicators (Ei) are independent variables whereas the
green environment indicator (Gi) is dependent variables. That is the green environment resilience is a
function of social and eco- efficiency indicators, which consists both environment and economical aspects. Standing from this notion, it is possible to formulate a linear relationship between these variables. Each variable also depends on own independent factors. This model formulation ultimately aimed at integrate and to show the relation of social aspect and eco efficiency with the green environment resilience. WBCSD (2009) and ESCAP (2011) proved that eco-efficiency, which consists economic and ecological aspects, could reduce environmental problem. What was left were social aspects integration into eco- efficiency
indicators and built a socio - eco efficiency framework. This framework was constructed, in this study, using instrumental variable model.
Hence, the following variable relation and model formulation proved that they have relation and association with green environment. First, let social aspects of people in industrial zone depend on factors. i.e social aspect is a function of factors (Xi) and other variables (Zi). Where, i is the number of factors in each
variable.
S =f (Xi, Zi) ……….……… …..……(1)
Where;
Xi is consisted of factors, which explained the social aspect of people like socio- demographic
characters, consumption behavior, culture and perception, health and etc.
Zi are factors influenced the social aspect includes water price and quantity consumed, lack
accessibility of infrastructure services, pollution, and depletion of resources like groundwater, behaviour, norms, habits and etc. Thus, social aspects linear function is explained as
Si = a1+b2Xi+c1Zi+ui……….……….…(2)
This indicates that social aspect is a function of industry’s product, resource consumption (Xi) and other
factors (zi) due to industrialization process. Where, ui is error term which may found in the process of data
survey or analysis stage.
The Eco- efficiency indicators applications were assumed varying and depending across the people and factory’s consumption and production activity. This study assumed and proposed that eco-efficiency application is determined by society’s progress in and outside the industrial zone. Therefore, eco efficiency is a function of social aspects in and outside the factory (Si) and including other factors (Ri) such as types of
factory consumption activity, technology and green job searches used to reduce an environmental pollution and etc. That is eco efficiency indicator application (Ei) is explained as;
Ei= f(Si, Ri)………..……….………….…………(3)
Ei = a2+b2Si+C2R+u2………..………..………….…….……..………(4)
Third, the next relation is built between green environment (Gi), eco efficiency (Ei) and other factors (Yi).
This is standing from the notion that green environment is depending on eco efficiency and social aspects as well as other factors (Yi) such as factory’s consumption and production activities. Household’s water
quantity consumption and recycling relation becomes: -
Gi = f(Ei,Yi) ……….………..……….(5)
Whereas, in a linear form:
Gi= a3+b3Ei+c3Yi+u3……… ………...……….………..…………(6)
Substitute equation (2) into equation (4) and insert equation (4) in to equation (6), we get Gi= a3+b3(a2 + b2a1 +b2b1Xi + b2c1Z1 + c2Ri) + c3Yi+vi, in simplified way
Gi= (a3+b3a2+b3b2a1) + b3b2b1Xi + b3b2c1Zi + b3c2Ri + C3Yi + vi ………..……….(7)
Suppose that α = a3+b3a2 + b3b2a1, β = b3b2b1, θ = b3b2c1 and λ = b3c2. Substitute these variables in
equation (7), we get a linear regression model, which describe green environment, depends on social
aspects, eco efficiency and other factors including errors.
Gi=α+βiXi+ θiZi+ λRi+ C3Yi + vi……….………..………...8
Equation (5) is the reduced form of the structured equation. Along similar calculation, let b1 = β1/(b3b2), b2
= β1/(b3b1), b3 = β1/(b2b1), c1 = θ/(b3b2) and c2 = λ/b3
Equation 1, 2 and 3 used and helped to estimate the parameters or value of coefficients. Thus, Gi depends
determined by the joint interaction effect of the social aspects and eco efficiency indicators called socio- eco efficiency, which is a contribution of this study.
Scholars discussed in problem statement and literatures were ignored the social aspects while they investigated the environmental problems by using eco-efficiency indicators in the production life cycle of a product. As argued so far, social aspects are found outside the model and hence this study incorporated social aspects in and outside the factory to get socio- eco efficiency frameworks. This study instrumental variable model proved exogenity of social indicators. It also estimated the predicted value of eco efficiency and social aspects using equation (4) in the first stage regression. However, instrumental variable model (IVM) would have its own limitation to estimate the value of estimator’s equation (8) in the first stge. Two Stage Least Squares estimation (TSLS), therefore, applied to determine social indictor’s effect on consumption and green environment tradeoffs. Indicators in the model were supposed to be endogenous and exogenous variables respectively.
3.8.3.1 Endogenous and Exogenous Factors
This study used both endogenous and exogenous factors to build socio- eco efficiency framework which balance tradeoffs between consumption growth and green environment. Based on Guajarati (1983 &2004); Greene (2011) and Wooldridge (2012), econometric use the terminology “Endogenous” means “determined within the system.” That is, a variable is jointly determined within the model subject to simultaneous causality. Whereas, exogenous variables are not determined in the model but have impact to influence the dependent variables. All part of exogenous factors could not influence the explained factors. Instead, some part of exogenous variables, which is associated with explanatory factors, have some bearing on the explained factors. In the context of this study, endogenous variables were eco-efficiency indicators, which interrelated with the residuals, and determined in the model. In other words, consumer’s economic and environmental indicators endogenously influenced tradeoffs between consumption growth and green environment problems.
Consumer’s social aspects (consumption culture), however, were exogenous variables which, are partly associated with eco efficiency indicators and have indirect impact on tradeoffs between consumption and green environment. In other words, social aspects are not determined in the system and uncorrelated with
the error term (ei). However, they are associated with eco-efficiency indicators which consists both
economic and environmental issues. This interpretation is narrow and hence instrumental variable regression was used to address omitted variable bias and errors-in-variable bias but not just simultaneous causality bias. Precisely, an endogenous variable is correlated with error terms (ei) whereas exogenous is
uncorrelated with error terms (ei).
Step one
i. Exogenous Factors (Social aspects, Si)
Instrumental variable model regression, loosely, breaks eco efficiency indicators into two parts. A part that might be correlated with ei, and a part that is not. By isolating the part that is not correlated with residuals
(ei), it is possible to estimate coefficients (parameters). To attain this, instrumental variable should be valid.
Hence, it is assumed that instrument relevance is exist when the covariance of instrumental and independent variables Cov(Si,Ei) and instrument exogeneity Cov(Si, ei) would be equal to zero.
Step Two: Model Justification
One of the basic justifications and rationality to apply multiple linear regression models is to integrate instrumental variable (social aspects) to eco efficiency and consists of Xi’s number of endogenous variables
determined in the model. Accordingly, Greene (2011) and Guajarati (2004) assumptions helps to explain important threats to internal validity. That is omitted variable bias from a variable that is correlated with Ei
but is unobserved cannot be included in the regression. Whereas, simultaneous causality bias endogenous explanatory variables assumed: (Ei causes Gi, Gi Causes Ei) and Errors in variables bias (Ei is measured
Gi.).
Step three: Factors in the Model
According to the given assumption in step two, suppose that Gi represents green environment resiliency
(dependent variable), Ei consists of various eco- efficiency indicators (explanatory variables), Si consists of
Step four: Multiple Linear Regression Model
Suppose that green environmental resilience is depending up on eco efficiency indicators and social aspect in growing industrial zones. That is in a function form:
Gi = f(Ei, Si)………...………..……(1)
Where;
Gi = green environment resilience
Ei = eco efficiency
Si = social aspects such as culture, norms, habits, and etc.
In a linear regression form:
Gi=β0 + β1iEi + ei……….………..……….…………..………(2)
Step five: Assumptions
In order for a variable Sito serve as a valid instrument for Ei, first the model consists of m endogenous (eco
efficiency indicators) and k number of exogenous variables (social aspects). Second, the instrument (social aspect) must be determined outside the model. That means only eco - efficiency is investigated within the model to reduce environmental problems but social aspect is not considered during consumption process. In other words, Cov (Si,ei) = 0. Third, the instruments, social aspect (Si) were correlated with endogenous
explanatory variable (eco efficient indicators (Ei)). That is Cov (Si, Ei) ≠ 0.
Step six: Estimation and Interpretation of Parameters
The instrumental variable regression breaks the E parts in two parts as explained so far. Hence, it detects movements in Ei that are uncorrelated with ei, and uses to estimate coefficients (βi). To find the value of
First stage: isolate the part of Ei uncorrelated with the residuals (ei) but correlated with Si. Regress Ei on Si
using Ordinary Least Square Techniques (OLS). That is
Ei= f(Si)……….….………..…….(3)
From this function, it is possible to formulate linear regression between eco efficiency indicator (Ei) and
social aspects (Si).
Ei= α0 +α1Si+ei …………...(4)
Since Si is uncorrelated with ei in equation (2) and also α1Si+ui is uncorrelated with ei, αi’s are estimators of
Si and their value will estimate after data survey.
Meanwhile, this proposal will compute the predicted value of Ei, which is;
Ei=α0 +α1Si ; where, I = 1, 2,……….………..…...….……….…(5)
The predicted value of the estimator or coefficients in equation (5) will tell us the directional change and association between eco efficiency indicators and social aspect. Nevertheless, it does not predict the estimator of predicted Ei. Thus, this study passed in the following steps to find the solutions.
Second Stage: to compute the predicted estimator values of eco efficiency indicators (i) in the interest of
green environmental resilience (Gi), replace the value of Ei by its prediction, Si. Such that this proposal
regress Gi on using OLS to get the estimators of βi’s and explore the association between the dependent
and explanatory variables. That is
Gi = β0 + β1Xi + ei………...………..………….….……….……….(6)
The resulting estimator of equation (6), which is βi’s the two stage least square estimator (TSLS) or βiTSLS.
These estimators will show how and how much the predicted value of eco efficiency indicator variables determine or changes the green environmental resiliency.
Step six: Testing the Model using Wu-Hausman Test
To test the endogenity and exogenity of the variables in the model, this study applied the idea of Hausman test which help to see if the estimates from OLS and IV are different. If this problem will come, the proposal will use auxiliary regression which is easiest way to do this test. Hausman (1978) and in Guajarati (2004) compares the OLS and TSLS estimates and determining whether the differences are significant. If they differ significantly, it was concluded that Ei is an endogenous variable. This would be achieved by
estimating the first stage regression:
Ei=α0 +α1Si+ui………....…… ………..………...….(7)
Assume that, since, each instrument is uncorrelated with ei, Ei is uncorrelated with ei only if ui is
uncorrelated with ei. To test this, this study formulated and ran the following regression using OLS
methods:
Gi = β0 + β1iEi + θi +ei….………...…..………(8)
Test whether θ = 0 using standard t-test that is If θ = 0, null hypothesis
if θ ≠ 0 alternative hypothesis
Thus, the result would be concluded by rejecting the null hypothesis; it is possible to say that Ei is
endogenous variables, since ui and ei are correlated. With the same procedures, exogenity of variables Si