SMOKING: This is measured as dummy variable equal to 1 if the individual is
having smoking habit, otherwise 0.
DRINKING: This is measured as dummy variable equal to 1 if the individual is
having alcohol drinking habit, otherwise 0.
PerCapitaAssets: This is measured in Indian rupees (INR).
SEX: This is measured as dummy variable equal to 1 for male and 0 for female.
AGE: Age of the individual is measured in number of years.
EDUCATION: This is coded as follows: 1 = Illiterate; 2 = below primary;
3 = Primary; 4 = Middle; 5 = Secondary/Metric; 6 = Technical; 7 = Graduate;
8 = Post graduate and above.
OCCUPATION: This is measured as dummy variable equal to 1 if the individual is
in the occupation of self-farming or agricultural laborer, 0 otherwise.
Note that the dependent variable in Eq. (9) is a censored variable, i.e., the
dependent variable is zero for corresponding known values of independent variables for part of the sample. Therefore, we use Tobit model for estimating the
demand for mitigating activities:
mS i ¼ a þ bx i þ u i if RHS [ 0
0
otherwise
ð11Þ
where mS i refers to the probability of the ith individual incurring positive medical
expenditure and x i denotes a vector of individual characteristics, such as assets, age,
sex, education, pollution parameter, etc.
In Eq. (10) the dependent variable is a count of the total number of workdays
lost due to air pollution-related illness by an individual during the particular period;
therefore, there are zeros for many observations. In this case, Poisson regression
model is appropriate as it considers the predominance of zeros and the small values
and the discrete nature of the dependent variable. The least square and other linear
regression models do not take into account these features. The Poisson regression
model can be stated as follows:
probðY i ¼ y i =x i Þ ¼ l
y i
i e
Àl i =y i ; y i ¼ 0; 1; 2; . . .:
ð12Þ
This equation is nonlinear in parameters; therefore, for estimation purpose by
taking its natural log we convert it into an equation which is linear in parameters.
Note that the Poisson regression model is restrictive in many ways. For example, the
assumption that the conditional mean and variance of y i , given x i are equal is very
strong and fails to account for over dispersion.
5 The data used in the estimation of
Eqs. 9 and 10 and the background information are provided in Tables 2, 3, and 4.
5
Similar estimation procedure is followed by Gupta (2008).
Economic Impact of Air Pollution from Agricultural …
305
having smoking habit, otherwise 0.
DRINKING: This is measured as dummy variable equal to 1 if the individual is
having alcohol drinking habit, otherwise 0.
PerCapitaAssets: This is measured in Indian rupees (INR).
SEX: This is measured as dummy variable equal to 1 for male and 0 for female.
AGE: Age of the individual is measured in number of years.
EDUCATION: This is coded as follows: 1 = Illiterate; 2 = below primary;
3 = Primary; 4 = Middle; 5 = Secondary/Metric; 6 = Technical; 7 = Graduate;
8 = Post graduate and above.
OCCUPATION: This is measured as dummy variable equal to 1 if the individual is
in the occupation of self-farming or agricultural laborer, 0 otherwise.
Note that the dependent variable in Eq. (9) is a censored variable, i.e., the
dependent variable is zero for corresponding known values of independent variables for part of the sample. Therefore, we use Tobit model for estimating the
demand for mitigating activities:
mS i ¼ a þ bx i þ u i if RHS [ 0
0
otherwise
ð11Þ
where mS i refers to the probability of the ith individual incurring positive medical
expenditure and x i denotes a vector of individual characteristics, such as assets, age,
sex, education, pollution parameter, etc.
In Eq. (10) the dependent variable is a count of the total number of workdays
lost due to air pollution-related illness by an individual during the particular period;
therefore, there are zeros for many observations. In this case, Poisson regression
model is appropriate as it considers the predominance of zeros and the small values
and the discrete nature of the dependent variable. The least square and other linear
regression models do not take into account these features. The Poisson regression
model can be stated as follows:
probðY i ¼ y i =x i Þ ¼ l
y i
i e
Àl i =y i ; y i ¼ 0; 1; 2; . . .:
ð12Þ
This equation is nonlinear in parameters; therefore, for estimation purpose by
taking its natural log we convert it into an equation which is linear in parameters.
Note that the Poisson regression model is restrictive in many ways. For example, the
assumption that the conditional mean and variance of y i , given x i are equal is very
strong and fails to account for over dispersion.
5 The data used in the estimation of
Eqs. 9 and 10 and the background information are provided in Tables 2, 3, and 4.
5
Similar estimation procedure is followed by Gupta (2008).
Economic Impact of Air Pollution from Agricultural …
305