yti = Xtiβ + uti i = 1,.... N ; t = 1,..........., T
ut = μ + ε t (3)
ε t = λWε t + vt
Here uti is the regression disturbance. In a vector form, the disturbance vector of (3) is
assumed to have random parish effects as well as spatially autocorrelated remainder
disturbances. denotes the vector of random parish effects which are assumed to be IIN.
is the scalar spatial autoregressive coefficient with | | <1. W is a known N X N spatial
weight matrix whose diagonal elements are zero.
Data
The dataset used is the same as the one used by Paudel et al. except for the social capital.
The disaggregated nature of the water pollution data used in our study is a first attempt to
study whether previous aggregated findings with the EKC hold for Louisiana. We used
data on nitrogen, phosphorus, and dissolved oxygen concentration in water from each
watershed collected by the Louisiana Department of Environmental Quality. The pooled
data consisted of observations from 1985 to 1999 for 53 parishes in Louisiana.
We focused on three kinds of ambient quality data for conventional pollutants:
dissolved oxygen (DO), phosphorus (P), and nitrogen (N). DO is a direct indicator of
water quality. Contamination of watersheds by human sewage or industrial discharges
increases the demand for dissolved oxygen, resulting in less oxygen for fish and other
forms of aquatic life. At a considerably high level of contamination, one would expect
that fish populations start to decline because of pollution. A similar problem may arise
when water is enriched with nutrients such as nitrogen and phosphorus through runoff
and leachates from intensively fertilized agricultural areas (Grossman and Krueger,
13