countries at a high level of per capita GDP (such as the United States and Canada) are excluded from the analysis,
the cubic term is not significant; and (3) the N-shaped relationship is the result of data aggregation and thus
does not describe the relationship between GDP and CO 2 emissions in any single country or subset of countries.
The first criticism is highly relevant in regard to the analysis presented here. The coefficients presented in Table
1 suggest an upward turning point of $32,000 which is very near the upward limit of observed values (only 1
percent of observations with a value greater than this amount). However, even when observations greater than
$32,000 or select high-income countries such as the United States, Canada, and Norway are excluded from the
analysis, the cubic functional form continues to fit the data well. The N-shaped relationship uncovered in this
analysis does not, therefore, appear to result from the presence of a few extreme outliers. Nor does it appear to be
the result of data aggregation. Table 2, which displays statistics on the functional form that best fits the data from
individual countries with a per capita GDP greater than $12,000 US dollars as of 2004, shows that a quadratic,
downward-U form fit the data in over half of the cases (53.1 percent). A cubic, N-shaped relationship fit the
data, however, in more than one-third of the cases (37.5 percent). In short, the fact that an N-shaped relationship
adequately describes the relationship between GDP and CO 2 emissions in a sizable number of high-income
countries demonstrates that the functional form in question is not simply the result of data aggregation.
What do we need to know: What are the major sociological research questions?
One question that remains to be answered is, does the observed N-shaped relationship between GDP and carbon
dioxide emissions hold once other covariates have been included in the equation? Existing studies have suggested
that the following factors may explain why CO 2 emissions decline once a certain income threshold has been met:
(1) A sectoral shift from manufacturing to services. The argument here is that the shift from a manufacturing
to a service economy results in an improvement of environmental quality since production within an economy
dominated by services is likely to be less energy- and material- intensive than production within an economy
based on manufacturing. Because the service sector tends to be larger where incomes are higher, the transition
from manufacturing to services is expected to lead to reduced emissions of carbon dioxide at higher income
levels.
(2) Urbanization. A number of researchers argue that urbanization is associated with a net reduction in CO 2
emissions on the grounds that spatial concentration leads to the more efficient delivery of goods and services in
general and reduced transportation costs in particular. Because economic development is itself strongly associated
with urbanization, any decline in emissions at higher income levels could be based at least in part on the fact that
higher-income countries tend to be more urbanized than lower-income ones
(3) Trade. Proponents of the “pollution haven” hypothesis suggest that any observed reductions in
environmental degradation at higher income levels are due not to structural changes in the economy or society
but rather to the movement of pollution-intensive production from higher- to lower-income countries. According
to this perspective, observed declines in emissions at higher income levels are illusory in the sense that they are
a reflection not of real reductions in emissions but rather the spatial relocation of polluters from high- to low-
income countries, leaving aggregate emissions unchanged.
Table 3 presents parameter estimates from a fixed-effects model which includes measures of sector size
(agriculture and service measured in terms of value added as a percentage of GDP, with manufacturing as the
excluded category), imports and exports (measured in terms of a percentage of GDP), and urbanization (measured
in terms of percentage of the population living in urban areas). Spatial constraints prevent me from discussing all