20
J. Swart and L. Brinkmann
trend reverses when a certain income threshold is reached. The turning points for
income per capita lie between 48,585 and 66,810 constant 2000 Brazilian Reais, thus
well within the range (cf. Table 10).
Included individually, the coefficient on population density is statistically significant with a positive sign, confirming our expectation that deforestation is more acute
in densely populated areas. Nevertheless, the estimate turns out to be insignificant
when including additional control variables.
4.3 Forest Fires
Table 3 reports the first differences estimates for our analysis on the relationship
between forest fires and economic complexity in Brazilian states and the federal
district between 2002 and 2009. As in Sects. 4.1 and 4.2, neither the ECI nor its
square term are significantly different from zero in the baseline specification. The
ECI estimate turns significant in specifications (6), (10), (11) and (12), suggesting
a linear and positive correlation between forest fires and economic complexity. In
other words, the number of forest fires rises with economic complexity. This can
be ascribed to the fact that rising economic complexity is likely to call for a more
efficient use of the land. Forest fires could be an easy way to clear the land for new
activities.
Further, there is robust evidence for an inverted U-shape relationship between
forest fires and per capita income. The square term of GDP per capita is significant
in all specifications and bears a negative sign. In line with the EKC hypothesis, this
implies that the number of observed forest fires in Brazilian states first increases with
rising incomes, but levels off and starts declining at a certain GDP per capita value.
This is because economic development is accompanied by a growing demand for
environmental quality, which will render the regulatory framework limiting forest
fires more effective.
Additionally, forest fires are significantly and negatively associated with trade
openness, a finding that stands firm to the inclusion of multiple variables. All
things equal, the number of annually observed heat sources in Brazilian states hence
decreases with intensified trade. There is also evidence that the number of forest fires
decreases with more hectares planted. While the sign of this estimate seems odd at
first sight, if plantation accounts for reforestation as well, this might be an indication that Brazil’s current sustainable reforestation projects are yielding the desired
outcomes (cf. Goldenberg and Roberts 2015).
In order to test the robustness of our findings, we run regressions (1)–(12) on a
subsample which does not comprise the state of São Paulo (unpublished results). São
Paulo accounts for ECI values more than ten times higher than the other Brazilian
states. Excluding this state confirms our findings from the full sample analysis with
respect to economic complexity and trade openness. In the subsample analysis forest
fires increase linearly with GDP per capita. This finding is not surprising when the
J. Swart and L. Brinkmann
trend reverses when a certain income threshold is reached. The turning points for
income per capita lie between 48,585 and 66,810 constant 2000 Brazilian Reais, thus
well within the range (cf. Table 10).
Included individually, the coefficient on population density is statistically significant with a positive sign, confirming our expectation that deforestation is more acute
in densely populated areas. Nevertheless, the estimate turns out to be insignificant
when including additional control variables.
4.3 Forest Fires
Table 3 reports the first differences estimates for our analysis on the relationship
between forest fires and economic complexity in Brazilian states and the federal
district between 2002 and 2009. As in Sects. 4.1 and 4.2, neither the ECI nor its
square term are significantly different from zero in the baseline specification. The
ECI estimate turns significant in specifications (6), (10), (11) and (12), suggesting
a linear and positive correlation between forest fires and economic complexity. In
other words, the number of forest fires rises with economic complexity. This can
be ascribed to the fact that rising economic complexity is likely to call for a more
efficient use of the land. Forest fires could be an easy way to clear the land for new
activities.
Further, there is robust evidence for an inverted U-shape relationship between
forest fires and per capita income. The square term of GDP per capita is significant
in all specifications and bears a negative sign. In line with the EKC hypothesis, this
implies that the number of observed forest fires in Brazilian states first increases with
rising incomes, but levels off and starts declining at a certain GDP per capita value.
This is because economic development is accompanied by a growing demand for
environmental quality, which will render the regulatory framework limiting forest
fires more effective.
Additionally, forest fires are significantly and negatively associated with trade
openness, a finding that stands firm to the inclusion of multiple variables. All
things equal, the number of annually observed heat sources in Brazilian states hence
decreases with intensified trade. There is also evidence that the number of forest fires
decreases with more hectares planted. While the sign of this estimate seems odd at
first sight, if plantation accounts for reforestation as well, this might be an indication that Brazil’s current sustainable reforestation projects are yielding the desired
outcomes (cf. Goldenberg and Roberts 2015).
In order to test the robustness of our findings, we run regressions (1)–(12) on a
subsample which does not comprise the state of São Paulo (unpublished results). São
Paulo accounts for ECI values more than ten times higher than the other Brazilian
states. Excluding this state confirms our findings from the full sample analysis with
respect to economic complexity and trade openness. In the subsample analysis forest
fires increase linearly with GDP per capita. This finding is not surprising when the