Economic Complexity and the Environment: Evidence from Brazil
7
complexity in 1995, the country ranked 54th in 2014 (Observatory of Economic Complexity 2017). Interestingly, while Brazil’s current economic and political crisis has
stopped this positive growth trend altogether, the country has seen its position in the
global economic complexity ranking strengthen simultaneously.
Alongside the prosperity brought about by Brazil opening up to international trade,
severe environmental problems emanated during this period. The country’s most
notorious environmental problem is the deforestation of its rain forests. Deforestation
in Brazil has been mainly driven by the expansion of agriculture. Hundred hectares
of the Amazon forest have been cut or burnt down each year for the large-scale
settlement of farmlands designated for cocoa, coffee, soybeans, sugarcane and cattle
pastures. By no means exclusive to the Amazon basin, these practices have also
reached other regions in Brazil (e.g. the tropical savanna of the Cerrado).
Moreover, urbanisation rates have risen dramatically in the last two decades.
According to the World Bank (2017), approximately 85% of the more than 200
million inhabitants in Brazil lived in urban areas in 2015. Particularly in the economic
hubs of São Paulo, Brazil’s largest city and one of the largest urban areas worldwide,
Rio de Janeiro and Belo Horizonte, air pollution as well as waste generation represent
serious environmental problems.
The paper is structured as follows. Section 2 briefly reviews the EKC literature,
first summarising the findings on the world or specific subsets of countries, and
secondly summarising Brazil-specific findings. Section 3 describes the proposed
model, the data used and the methodology chosen. Section 4 presents the empirical
results for four environmental degradation indicators analysed: Section 4.1 for solid
waste generation, Sect. 4.2 for deforestation, Sect. 4.3 for forest fires, and Sect. 4.4
for air pollution. Section 5 concludes by outlining the implications of our results and
the possible scope for improvement and extension.
2 Literature Review
The EKC has its roots on the seminal work by Grossman and Krueger (1991), the
first authors to explicitly find evidence for an inverted U-shaped relationship between
indicators of environmental degradation and income. Following their work other
empirical papers found evidence for an EKC (see e.g. Shafik and Bandyopadhyay
1992; Panayotou 1993; and Selden and Song 1994). Notwithstanding the empirical
evidence of the first EKC studies, Stern et al. (1996) examine the concept of the EKC
critically on both theoretical and empirical grounds. In a later review, Stern (2004)
argues that most of the EKC literature of the 1990s is empirically weak, as it fails
to take into account statistical properties likely to influence the estimation results
(such as heteroskedasticity, serial correlation and cointegration), and the possible
existence of simultaneity and omitted variables bias. Furthermore, the coefficients
estimated in the first EKC studies differ substantially depending on the subsample
used. Stern (2004) notes that no or only weak evidence for the EKC hypothesis can be
found when accounting for diagnostic statistics, using appropriate techniques, and
7
complexity in 1995, the country ranked 54th in 2014 (Observatory of Economic Complexity 2017). Interestingly, while Brazil’s current economic and political crisis has
stopped this positive growth trend altogether, the country has seen its position in the
global economic complexity ranking strengthen simultaneously.
Alongside the prosperity brought about by Brazil opening up to international trade,
severe environmental problems emanated during this period. The country’s most
notorious environmental problem is the deforestation of its rain forests. Deforestation
in Brazil has been mainly driven by the expansion of agriculture. Hundred hectares
of the Amazon forest have been cut or burnt down each year for the large-scale
settlement of farmlands designated for cocoa, coffee, soybeans, sugarcane and cattle
pastures. By no means exclusive to the Amazon basin, these practices have also
reached other regions in Brazil (e.g. the tropical savanna of the Cerrado).
Moreover, urbanisation rates have risen dramatically in the last two decades.
According to the World Bank (2017), approximately 85% of the more than 200
million inhabitants in Brazil lived in urban areas in 2015. Particularly in the economic
hubs of São Paulo, Brazil’s largest city and one of the largest urban areas worldwide,
Rio de Janeiro and Belo Horizonte, air pollution as well as waste generation represent
serious environmental problems.
The paper is structured as follows. Section 2 briefly reviews the EKC literature,
first summarising the findings on the world or specific subsets of countries, and
secondly summarising Brazil-specific findings. Section 3 describes the proposed
model, the data used and the methodology chosen. Section 4 presents the empirical
results for four environmental degradation indicators analysed: Section 4.1 for solid
waste generation, Sect. 4.2 for deforestation, Sect. 4.3 for forest fires, and Sect. 4.4
for air pollution. Section 5 concludes by outlining the implications of our results and
the possible scope for improvement and extension.
2 Literature Review
The EKC has its roots on the seminal work by Grossman and Krueger (1991), the
first authors to explicitly find evidence for an inverted U-shaped relationship between
indicators of environmental degradation and income. Following their work other
empirical papers found evidence for an EKC (see e.g. Shafik and Bandyopadhyay
1992; Panayotou 1993; and Selden and Song 1994). Notwithstanding the empirical
evidence of the first EKC studies, Stern et al. (1996) examine the concept of the EKC
critically on both theoretical and empirical grounds. In a later review, Stern (2004)
argues that most of the EKC literature of the 1990s is empirically weak, as it fails
to take into account statistical properties likely to influence the estimation results
(such as heteroskedasticity, serial correlation and cointegration), and the possible
existence of simultaneity and omitted variables bias. Furthermore, the coefficients
estimated in the first EKC studies differ substantially depending on the subsample
used. Stern (2004) notes that no or only weak evidence for the EKC hypothesis can be
found when accounting for diagnostic statistics, using appropriate techniques, and