jth variable for the ith country given j equal to variable used in the regression. Social
capital index is calculated by dividing each SC value thus calculated with the highest SC
value. Therefore SCINDEX ranges from 0 to 1.
Panel data regression
Data that have both time series and cross sections, usually referred to as panel data, are
common in economics. Many recent studies of the Kuznets curve have used panel data
because it provides a rich source of information about the economy and allows
researchers great flexibility in modeling differences in behavior across individuals. In our
study, we used panel data covering three different water pollutants in 53 Louisiana
parishes over a 14-year period.
Kuznets curve models have been estimated either in quadratic or in cubic
specifications between pollutant concentration and per capita income. We adopt both of
these specifications in our analysis. The general form of the panel data model used to
describe the relationship between pollution and income in this study is given in equation
(2).
m
Pt = a +∑ fikSCi, + βm. Dt + u„ (2)
k=1
Here, p is a water pollutant (nitrogen, phosphorus or dissolved oxygen), SC is per capita
income, i and t represent indices of parish and time, respectively. Population density
(persons per square mile) is accounted by D. We estimated the model with quadratic and
cubic specifications so m=2 when income pollution relationship is specified as quadratic
and m=3 if income pollution relationship is specified as cubic. Population density is used
in the model as a proxy for human behavior on water pollution. The hypothesis
11