12
Hence, seven countries (the U.S., Turkey, Indonesia, Malaysia, Japan, Saudi Arabia, and Russia)
have several MTI scores – one for each of their coastlines - while having only one national value of
independent variables to correspond to them. This is also the case for sixty-seven island-colonies,
where MTI values are available but there are no corresponding values of the independent variables.
We therefore chose to exclude these cases from the analysis. In doing so, considerable variance in
our dependent variable is lost, but we still consider our strategy of excluding cases a safer option
than alternative approaches. An alternative strategy would have been to average the values of MTI
for countries with several coastlines in order to obtain a single national score for the dependent
variable to correspond with other variables. Another strategy would have been to impute data for
independent variables to the regions or islands-colonies with no regional measures. However, both
of these other strategies have obvious problems. The strategy of creating average values of MTI for
coastal regions would distort the data. The strategy of imputing data for the coastal regions or col-
onies, might not correspond to reality and may thus produce misleading results.
The results presented in the next section follow the model described above. However, we
also performed a number of alternative estimations. We tested several lag structures. Using differ-
ent lags of the independent variables in time indicated that the one-year lag produced the most
significant results. Since previous studies found a U-shaped relationship between GDP and envi-
ronmental outcomes (e.g., Grossman and Kreuger, 1993, 1995) as well as between democratic de-
velopment and environment (e.g., Buitenzorgy and Ancev, 2011), we also tried a similar model but
with squared values of those variables included. However, the results were similar to those present-
ed in the tables. Granger causality testing seems to confirm that no reversed causality exists be-
tween our dependent and independent variables.
Results and analysis
In this section we empirically explore the relationship between levels of democracy and annual
changes in the marine trophic index during different stages of economic development. We first
apply our equation to the whole sample to investigate the relationship between our variables of
interest on the global scale and across time. In order to find out whether democracy exerts an influ-
ence on the changes in marine trophic levels during different stages of economic development, we
then explore this relationship in different income groups.
13
Table 1 presents the results from our multivariate model on the global sample over all
available years. The unit of analysis is country-year and the sample includes 142 marine coastal
states over the years 1972-2006. The analysis shows that democracy is significantly and negatively
correlated with changes in marine trophic levels. According to this pattern, less democratic coun-
tries tend to have less healthy marine ecosystems. However, when we proceed to divide countries
based on their income, we can note some more detailed trends, not visible in the first analysis.
TABLE 1. THE INFLUENCE OF DEMOCRACY ON CHANGES IN MARINE TROPHIC LEVELS
DV: Differenced MTI
Model 1
Model 2
Model 3
Model 4
Model 5
Democracy
-0.00220**
-0.00226**
-0.00301***
-0.00298***
-0.00269**
(0.000792)
(0.000774)
(0.000854)
(0.000845)
(0.000907)
Openness to trade
0.00218
0.000177
0.00111
-0.00334
(0.00505)
(0.00480)
(0.00590)
(0.00715)
Population
0.0196*
0.0199*
0.0210*
(0.00779)
(0.00789)
(0.00818)
GDP per capita
-0.00329
-0.00345
(0.00741)
(0.00660)
Trawling intensity
0.00314
(0.00290)
Constant
0.0122**
0.00338
-0.284*
-0.263*
-0.251*
(0.00447)
(0.0213)
(0.123)
(0.122)
(0.125)
Observations
4,255
4,133
4,100
4,100
4,015
R-squared
0.001
0.001
0.002
0.002
0.003
Number of countries
142
138
137
137
137
Robust standard errors in parentheses, *** p<0.001, ** p<0.01, * p<0.05. Groups are divided based on GNI per capita in
2010 constant US dollars. All the independent variables are lagged 1 year. Openness to trade, population, GDP per capita and
trawling intensity are log-transformed.