Environmental Engineering Reference
In-Depth Information
ministic, i.e. future forecasts are highly problematic. We come back to this point later on,
when we discuss the implications of the EKC literature for LDCs. Even if an EKC rela-
tionship is found, there is the possibility of a second turning point. To check for this pos-
sibility, studies would need to add a cubic per capita GDP term to equation (3.1). Studies
that have found second turning points include De Bruyn and Opschoor (1997) and Binder
and Neumayer (2005).
Country-speci
fi
c
fi
xed and year-speci
fi
c time e
ff
ects are often required, but sometimes
not included. Country
ects are required if per capita GDP or some other explana-
tory variables are correlated with country-speci
fi
xed e
ff
c time-invariant factors, such as geo-
graphical factors (climate, land size and resource endowments - see Neumayer, 2002a;
2004), or institutional quality. Year-speci
fi
ects are required if there are global
changes in environmental indicators, perhaps due to global advances in technology, that
have a roughly equal impact on countries at any given point of time. Where country-
speci
fi
c time e
ff
ects and are con-
tingent to the sample at hand. Strictly speaking, no out-of-sample predictions are possible
for such estimation results.
If the environmental indicator and GDP per capita are both trending over time (in tech-
nical terms: are non-stationary), then spurious regression results are possible. Year-
speci
fi
c
fi
xed e
ff
ects are included, the results are conditional on these e
ff
fi
c time dummies mitigate, but do not solve, the problem. Estimating the model in
fi
erences might work as a solution. Co-integration is superior, but only if both vari-
ables are truly co-integrated. Very few studies have taken this potential problem seriously
(Galeotti et al., 2006; Perman and Stern, 2003; Stern, 2000; Stern and Common, 2001;
Wagner and Müller-Fürstenberger, 2005).
Where EKC exists, this could be partly due to a trade e
rst di
ff
ect, i.e. rich countries may have
become clean partly by importing products that are polluting in production from lower-
income countries. See our discussion below.
Even where EKCs exist, with median GDP per capita far below mean GDP per capita
the environmental implications can be unpleasant for many low-income countries for
many years to come (Cole and Neumayer, 2005).
Why is there a distinction between these three di
ff
erent groups of environmental indi-
cators? One possible explanation is that those that are very important to human health,
such as local public goods, are not easily externalized and tend to improve at low levels of
income, whereas those that are global public goods, and are quite easy to externalize to
others, such as CO
2
emissions, worsen with economic growth (Sha
ff
k, 1994, p. 768).
However, one of the key questions that academics have addressed since the EKC hypoth-
esis came under scrutiny is whether or not the EKC relationship is quasi-automatic or
policy induced (Grossman and Krueger, 1995), to which we now turn.
fi
EKC and policy
The reduced-form econometric models that have commonly been used in EKC studies do
not test the pro-growth hypotheses as discussed above (Grossman and Krueger, 1995,
p. 372). However, other studies have analyzed the factors that in
uence environmental
change on a more disaggregate level (Neumayer, 2003b, p. 84). Selden et al. (1999) ana-
lyzed scale, composition and technique e
fl
ects at the sector level to decompose changes in
various US emissions.
1
Scale is indicated by the growth of emissions when the ratio of
emissions to GDP remains constant. Composition e
ff
ff
ects are changes in emissions due to
