# Abstract:

This research aims to analyze the relationship between the deforestation of the Atlantic Forest and economic activity, under the hypothesis of the Environmental Kuznets Curve, considering the municipalities of the state of Ceará. For this, it is estimated a Tobit model for panel data, in the period from 2011 until 2017, considering the GDP per capita and controlling for associated factors with both deforestation and environmental protection. Unlike the expectations, the population density and cattle farming soften deforestation activity. The evidence found for the relationship between deforestation and GDP per capita rejects the “Inverted-U” hypothesis, and yes, suggests the “N” format, indicating that deforestation in the region may be cyclical. Therefore, there are no indications to ensure that the economic activity of the municipalities analyzed assure by itself the environmental protection and sustainable use of the Atlantic Forest and associated ecosystems in the state of Ceará.

Keywords:
Ceará; Environmental Kuznets Curve; deforestation; Atlantic Forest

# Resumo:

Palavras-chave:
Ceará; Curva de Kuznets Ambiental; desmatamento; Mata Atlântica

# 1. Introduction

The degradation of forest ecosystems is due to several factors, especially activities of anthropic origin. In this sense, there is a substantial concern that, as the economic development advances, the environmental pressure on ecosystems exceeds the capacity supported by the environment itself. The relationship between economic development and environmental degradation is found in the Environmental Kuznets Curve (EKC), initially proposed by Grossman & Krueger (1995)Grossman, G., & Krueger, A. (1995). Economic growth and the environment. The Quarterly Journal of Economics, 110(2), 353-377..

EKC's premise assumes that, at low levels of development, the growth of income per capita (pc) induces an increase in environmental degradation. However, after a certain level of income, this logic would reverse, with an increase in income pc leading to a reduction in environmental degradation, characterizing a curve in the shape of an “Inverted-U”. Thus, Grossman & Krueger (1995)Grossman, G., & Krueger, A. (1995). Economic growth and the environment. The Quarterly Journal of Economics, 110(2), 353-377. related income pc to water and air quality indicators and found evidence for several developed countries, that environmental degradation increases with economic growth to an inflection point from which there is an improvement in environmental quality resulting in the EKC.

In Brazil, several studies elaborate on the relationship between economic development and indicators of environmental degradation. Hence, one of the most pertinent ecological guidelines, concerns deforestation in national biomes, focusing on regions such as the Amazon and Cerrado (e.g. Colusso et al., 2012Colusso, M. V. S., Parré, J. L., & Almeida, E. S. (2012). Degradação ambiental e crescimento econômico: a curva de Kuznets ambiental para o cerrado. Revista de Economia e Agronegócio, 10, 335-357.; Oliveira et al., 2011Oliveira, R. C., Almeida, E., Freguclia, R. S., & Barreto, R. C. S. (2011). Desmatamento e crescimento econômico no Brasil: uma análise da curva de Kuznets ambiental para a Amazônia Legal. Revista de Economia e Sociologia Rural, 49(3), 709-739.) especially due to their territorial extension ecosystems, the expansion of the agricultural frontier in these places and the availability of information. On the other hand, there is a gap in the research that links economic development and changes in forests in other national biomes (i.e. Caatinga, Atlantic Forest, Pampa, and Pantanal).

Regarding the Atlantic Forest and natural non-forest areas1 1 Lowlands, natural altitude fields, vegetation refuges, dunes, herbaceous restinga, apicum, wetland and humid field. , their limits currently occupy 12.4% of the country, distributed in 17 states. According to the “Atlas of the Remnants of the Atlantic Forest”, from 1985 to 2019, the anthropic action deforested around 1.95 million acres of this forest coverage. Deforestation between 2018 to 2019 was 14,502 acres, 27.2% higher than the period between 2017 and 2018. Of the 17 states monitored between 2018 and 2019, nine registered zero deforestation (<100 acres), namely: Alagoas (AL), Ceará (CE), Espírito Santo (ES), Goiás (GO), Paraíba (PB), Pernambuco (PE), Rio de Janeiro (RJ), Rio Grande do Norte (RN), and São Paulo (SP) (SOS Atlantic Forest Foundation & National Institute for Space Research [INPE], 2020a).

Among these regions, the state of Ceará stands out, with remnants of the Atlantic Forest and non-forest natural areas occupying 64,064 and 125,519 acres, respectively, or 1.3% of the state. In the region, 61 municipalities have remnants of native forest, and therefore, by observing the deforestation data, it can be noticed that, since 2014, the state has recorded zero deforestation (<100 acres) of Atlantic Forest, with a subsequently 40% increase in deforestation in the periods 2017 to 2018 and 2018 to 2019. Concerning non-forest areas, in the period from 2018 to 2019 alone, 804 acres of herbaceous restinga were deforested (Fundação SOS Mata Atlântica, 2020aFundação SOS Mata Atlântica, & Instituto Nacional de Pesquisas Espaciais – INPE. (2020a). Atlas dos remanescentes florestais da Mata Atlântica: período 2018-2019 (pp. 61). São Paulo: INPE.).

In light of the abovementioned situation, this paper aims to analyze the relationship between environmental degradation and economic activity, as proposed by the EKC hypothesis, in the 61 municipalities of Ceará that have remnants of the Atlantic Forest and associated ecosystems, considering the deforestation indicator. Therefore, the goal is to investigate if there is an “Inverted-U” relationship between GDPpc and the annual deforested area, obtained after the “Aqui tem Mata?” Project data, for the period between 2011 and 2017, using a panel data model. In addition to the variable related to economic activity, there are control variables associated with: agriculture, consumption, demography, economic and tax information, the labor market, and geoenvironmental aspects were added.

Although several surveys estimate an EKC for Brazil based on several sustainable development indicators, none of them used the deforestation of the Atlantic Forest and non-forest natural areas as a proxy for environmental degradation. Also, EKC's empirical analyzes focus on states and municipalities that have limits on the Amazon and the Cerrado. Hence, the literature does not present evidence for the state of Ceará, regarding the Caatinga and Atlantic Forest domains. Therefore, this research contributes to the EKC literature for the Atlantic Forest biome.

To achieve the proposed objectives, this article is divided into five sections, including this introduction. The next one is dedicated to exposing the empirical literature on EKC, focusing on deforestation. The source and description of the data, as well as the empirical model, are presented in the third section. Subsequently, the results are exposed, discussed, and analyzed. Finally, the conclusions are presented.

# 2. Literature Review

## 2.1 Empirical literature on EKC focused on deforestation

To conduct this literature review, we decided to apply the descriptors “Environmental Kuznets Curve” + “Deforestation” on Google Scholar’s database. The criteria for the inclusion of the available articles were: publications in national or international journals, in English, Portuguese or Spanish, with publication date starting in 2010, selection of updated references, and preference for studies carried out in Brazilian biomes. Based on these selection procedures, the content of this article is evaluated considering the study area, period, methodology, and results.

Oliveira et al. (2011)Oliveira, R. C., Almeida, E., Freguclia, R. S., & Barreto, R. C. S. (2011). Desmatamento e crescimento econômico no Brasil: uma análise da curva de Kuznets ambiental para a Amazônia Legal. Revista de Economia e Sociologia Rural, 49(3), 709-739. investigated deforestation under the hypothesis of EKC in the municipalities of the Amazônia Legal from 2001 to 2006. To do so, they controlled for indicators related to agriculture, demography, economics, forestry, and mineral extraction. They found an “N-inverted” format for the EKC, hence, deforestation decreases at low levels of income pc, experiences a period of increase, and decreases again when a high level of income pc is reached.

Ferreira & Coelho (2015)Ferreira, M. D. P., & Coelho, A. B. (2015). Desmatamento recente nos estados da Amazônia legal: uma análise da contribuição dos preços agrícolas e das políticas governamentais. Revista de Economia e Sociologia Rural, 53(1), 93-108. explored how the prices of agricultural commodities, public inspection policies, and rural credit, have affected deforestation in the Amazônia Legal region from 1999 to 2011. The results obtained suggest that commodities prices and rural credit policies increase deforestation, while inspection policies have helped to mitigate deforestation.

Regarding the discussion between deforestation and regional development, Teixeira, Bertella & Almeida (2012)Teixeira, R. F. A. P., Bertella, M. A., & Almeida, L. T. (2012). Curva de Kuznets ambiental para o estado do mato grosso. Análise Econômica, 30(57), 313-337. investigated the relationship between deforestation levels and income growth for 139 municipalities in the state of Mato Grosso (MT). Using data from 2006, an EKC was estimated for deforestation pc in relation to income pc and its quadratic term, in addition to: cattle over pasture area, demographic density, wood extraction pc, and spatial effects. The results found led to the inference that the EKC follows an “Inverted-U” format, but when using a cubic term for income, economic growth would not be linked to the deforestation of the municipalities of Mato Grosso.

Colusso et al. (2012)Colusso, M. V. S., Parré, J. L., & Almeida, E. S. (2012). Degradação ambiental e crescimento econômico: a curva de Kuznets ambiental para o cerrado. Revista de Economia e Agronegócio, 10, 335-357. discuss the relationship between environmental degradation and economic activity, from the perspective of EKC, in the Brazilian Cerrado, for the year 2008. The dependent variable is set as the deforested area of Cerrado in the municipalities of the biome and, as explanatory variables, there are: GDPpc and its quadratic and cubic shapes, population density, planted area, and cattle herd. The tested hypothesis resulted that, in the first stage, the growth of income pc contributes to the reduction of deforestation in the region; however, the continuous increase in income causes deforestation to increase again.

For MATOPIBA2 2 Current Brazilian agricultural frontier, the MATOPIBA covers the Cerrado biome of the states of Maranhão, Tocantins, Piauí and Bahia, and accounts for a large part of the national production of grains and fibers (Empresa Brasileira de Agropecuária, 2020). , Barros & Stege (2019)Barros, P. H. B., & Stege, A. L. (2019). Deforestation and human development in the Brazilian agricultural frontier: an environmental Kuznets curve for MATOPIBA. Revista Brasileira de Estudos Regionais e Urbanos, 13(2), 161-182. investigated the existence of an EKC of the relationship between deforestation and the Human Development Index (HDI) in the 337 municipalities of the region in 2010. The Exploratory Analysis of Spatial Data (AEDE) combined with Spatial econometrics corroborated the “Inverted-U” format of the EKC. The turning point, at which development reaches its maximum, is at an HDI of 0.57, with 28.18% of the municipalities being below this value. In addition, controls related to the advance of the agricultural frontier induce deforestation in the region.

Although the aforementioned articles directly address deforestation and its relationship to economic growth, there are others that, even though do not address the subject directly, explain the issuance of greenhouse gases as a consequence of deforestation and which in turn are affected by economic growth3 3 Almeida & Lobato (2019), Biage & Almeida (2015) corroborate that the increase in economic activity increases deforestation via the agricultural sector with: burning in agriculture, extraction of forest resources, and pasture areas, which reduce CO2 absorption, increasing GHG emissions. (i.e. Almeida & Lobato, 2019Almeida, M. G., & Lobato, T. C. (2019). A curva de Kuznets ambiental para a região norte do Brasil entre os anos de 2002 a 2015. Economia & Região, 7(1), 7-25.; Biage & Almeida, 2015Biage, M., & Almeida, H. J. F. (2015). Desenvolvimento e impacto ambiental: uma análise da curva ambiental de Kuznets. Pesquisa e Planejamento Economico, 45(3), 505-556.; Carvalho & Almeida, 2010Carvalho, T. S., & Almeida, E. S. (2010). A hipótese da curva de Kuznets ambiental global: uma perspectiva econométrico-espacial. Estudos Econômicos, 40(3), 587-615.).

Carvalho & Almeida (2010)Carvalho, T. S., & Almeida, E. S. (2010). A hipótese da curva de Kuznets ambiental global: uma perspectiva econométrico-espacial. Estudos Econômicos, 40(3), 587-615. tested the EKC hypothesis in a sample of 187 countries for 2004. The dependent variable is set as CO2 emissions and as independent variables, there are: GDPpc and its quadratic and cubic forms, exports pc, consumption pc energy, and a dummy indicating the signatory countries to the Kyoto Protocol. The results corroborate the “Inverted- U” format, thus, economic growth reduces the environmental impact of emissions. The addition of the cubic term, however, results in an “N” shape curve, demonstrating that high levels of growth can increase CO2 emissions. Moreover, Kyoto Protocol signatory countries have significantly reduced pollutant emissions.

Biage & Almeida (2015)Biage, M., & Almeida, H. J. F. (2015). Desenvolvimento e impacto ambiental: uma análise da curva ambiental de Kuznets. Pesquisa e Planejamento Economico, 45(3), 505-556. evaluated the EKC hypothesis applied to a panel data format, analyzing the differences between CO2 emissions by countries, depending on socio-economic development. As a result, a relationship between GDPpc and the CO2pc emission with the EKC in an “N” format became evident. In addition, the results showed that GDPpc is the variable with the least impact on CO2 emissions and that the environmental impact grows, essentially, due to the development of economies (economic development, social development, and quality of life).

Almeida & Lobato (2019)Almeida, M. G., & Lobato, T. C. (2019). A curva de Kuznets ambiental para a região norte do Brasil entre os anos de 2002 a 2015. Economia & Região, 7(1), 7-25. approach the EKC discussion for the Northern region of Brazil from 2002 to 2015. The authors linked CO2 emissions to GDPpc, and although they do not directly address deforestation, they took into account that the CO2 emissions are a consequence of factors such as deforestation itself, being treated as a proxy variable. The research concluded that the EKC for the northern region of Brazil has a “U” shape, not corroborating the traditional EKC format.

Relevant studies for the development of this paper refer to the evaluation of deforestation, without the need to test the EKC hypothesis. Their importance relates to the evaluation of the determinants of deforestation in Brazilian biomes, or the development of environmental impact indicators, in order to identify causes of environmental degradation.

From another perspective, Delazeri (2016)Delazeri, L. M. M. (2016). Determinantes do desmatamento nos municípios do arco verde - Amazônia Legal: uma abordagem econométrica. Economia Ensaios, 30(2), 11-34. listed the causes of deforestation in the Arco Verde communities in the Amazônia Legal between 2008 and 2012. The result led to conclude that cattle production has a greater incidence in the levels of deforestation in the 49 municipalities that make up the region and that the expansion of soybean crops is not significant to explain deforestation.

In an investigation into the occurrence of environmental impact decoupling from Brazilian economic growth, Soares & Almeida (2018)Soares, L. R., & Almeida, L. T. (2018). Desacoplamento de impactos ambientais no Brasil. Revista Iberoamericana de Economía Ecológica, 28(2), 21-43. grouped 13 environmental pressure indicators, monitored from the 1990s to mid-2014, in the dimensions4 4 Water: biochemical oxygen demand (1990 to 2014), beach water quality (1992 to 2012). Atmosphere: anthropogenic GHG emissions (1990 to 2014), industrial consumption of ozone-depleting substances (1992 to 2013). Biodiversity (1992 to 2013): protected land areas, marine protection areas. Sanitation (1992 to 2011): access of the population to drinking water, sewerage and to the domestic garbage collection service. Land: deforestation of the Amazônia Legal (1990 to 2014), use of fertilizer (1992 to 2013), land use (1990 to 2011). : water, atmosphere, biodiversity, sanitation, and land. The research corroborated the hypothesis that economic growth causes environmental impact, including on the biodiversity that comprehends the Atlantic Forest, but with less impact on the land dimension.

Castelo, Adami, Almeida & Almeida (2018)Castelo, T. B., Adami, M., Almeida, C., & Almeida, O. T. (2018). Governos e mudanças nas políticas de combate ao desmatamento na Amazônia. Revista Iberoamericana de Economía Ecológica, 28(1), 125-148. evaluated environmental public policies of the federal government in the fight against deforestation in the Amazon, from a historical survey of the activity for a period of 15 years (2002 to 2016). The empirical analysis carried out for the state of Pará using a panel modeling showed that, given the increase of 100 bovine heads, the deforested area has increased by 0.2km2, since the environmental policy was quite significant for the period analyzed.

Briefly, Table 1 summarizes the studies that assess the relationship between economic growth and deforestation in Brazilian biomes under the hypothesis of EKC. In general, these empirical studies consider econometric approaches for cross-sectional data or in panel data format. In addition, much of the national research is concentrated in the Midwest, North, and MATOPIBA municipalities. Thus, through this review, it can be said that the present research contributes to the EKC literature based on the analysis of the relationship between economic activity in the state of Ceará and the deforestation of the forest remnants of the Atlantic Forest and non-forest natural areas.

Table 1
Summary of the works considered by EKC and/or Deforestation.

# 3. Methodology

## 3.1 Data

This research analyzes the annual deforestation of the forest remnants of Atlantic Forest and natural non-forest areas - DEF, in acres. These data are from the project “Aqui tem Mata?”, an application elaborated with data from the “Atlas of the Atlantic Forest”, from the Fundação SOS Mata Atlântica (2020b)Fundação SOS Mata Atlântica, & Instituto Nacional de Pesquisas Espaciais – INPE. (2020b). Aqui tem Mata? Retrieved in 2020, August 9, from https://aquitemmata.org.br/#/
https://aquitemmata.org.br/#/...
that presents graphs and interactive maps with updated information on the state of conservation of the forests, mangroves and restingas in all 3,429 Brazilian municipalities with remnants of the Atlantic Forest.

In the survey that analyzes the changes in forest cover, two limitations stand out: the presence of cloud cover, which impairs image processing, hence there may be areas not observed; and, the limitation of the mapping, which requires a minimum area of 3 acres, both for the detection of forest changes and for the identification of forest remnants. For monitoring purposes, areas with deforestation of fewer than 3 acres are marked as evidence of deforestation and will be observed again in new versions of the reports (Fundação SOS Mata Atlântica, 2020aFundação SOS Mata Atlântica, & Instituto Nacional de Pesquisas Espaciais – INPE. (2020a). Atlas dos remanescentes florestais da Mata Atlântica: período 2018-2019 (pp. 61). São Paulo: INPE.).

The research comprising the period from 2011 to 20175 5 During the preparation of this survey, the municipal GDP for the year 2018 had not been released, so there is a limitation on the use of data on deforestation for the year 2017. in the state of Ceará resulted in 61 municipalities monitored and distributed among five of the seven mesoregions of Ceará (Figure 1). Mesoregion 1 comprises 26 monitored municipalities (50.51% of the total), followed by mesoregion 2 with 20 municipalities (28%) and mesoregions 3, 5, and 7 together with 14 monitored municipalities (21.49%).

Figure 1
Municipalities in Ceará monitored by the “Aqui tem Mata?” Program. Source: Based on research data

The deforestation process might happen because of several activities, especially anthropic ones. Based on this, we used information from agriculture, consumption, demography, economic and tax, labor market, and geoenvironmental aspects, from different sources, as shown in Table 2.

Table 2
Description of variables and data sources.

The economy, population, and territory information is extracted from the Brazilian Institute of Geography and Statistics (IBGE) and includes GDPpc, its quadratic and cubic form, the Gross Added Value of Agriculture (GAVpcagro), the Value of Extractive Production (VEP), Population Density (POPD) and the number of cattle (CAT).

The number of workers employed in the sectors of Agriculture, Livestock and Related Services and Forest Production Emp.primary comes from the Annual List of Social Information (RAIS), of the Ministry of Labor (Brasil, 2017Brasil. Ministério do Trabalho – MTB. (2017). A Relação Anual de Informações Sociais - RAIS. Retrieved in 2020, August 9, from http://www.rais.gov.br/sitio/index.jsf
http://www.rais.gov.br/sitio/index.jsf...
). While Municipal Electricity Consumption (ECpc), is provided by Ente Nazionale per L’energia Elletrica (Ente Nazionale per L’energia Elletrica, 2017Ente Nazionale per L’energia Elletrica – ENEL. (2017). Companhia Energética do Estado do Ceará – COELCE. Retrieved in 2020, August 9, from https://www.enel.com.br/pt-ceara.html
https://www.enel.com.br/pt-ceara.html...
).

Tax information such as the Municipal Participation Fund (FPM), the Tax on Circulation of Goods and Services (ICMS), and Tax Revenue (RT) were extracted from the Public Sector’s Accounting and Tax Information System (Sistema de Informações Contábeis e Fiscais do Setor Público Brasileiro, 2017Sistema de Informações Contábeis e Fiscais do Setor Público Brasileiro – SICONFI. (2017). Contas anuais. Retrieved in 2020, August 9, from http://www.tesouro.fazenda.gov.br/contas-anuais
http://www.tesouro.fazenda.gov.br/contas...
), created by the National Treasury Secretariat (STN).

The total annual observed Rainfall (RAIN) is from the Cearense Foundation for Meteorology and Water Resources (Fundação Cearense de Meteorologia e Recursos Hídricos, 2020Fundação Cearense de Meteorologia e Recursos Hídricos – FUNCEME. (2020). Portal Hidrológico do Ceará. Retrieved in 2020, August 9, from http://www.hidro.ce.gov.br/municipios/chuvas-diarias
http://www.hidro.ce.gov.br/municipios/ch...
) and the delimitation of the Semi-Arid Region (SAR) in Brazil is from the Ministry of Integration (Brasil, 2018Brasil. Ministério da Integração – MI. (2018). Nova delimitação Semiárido. Retrieved in 2020, August 9, from https://www.gov.br/sudene/images/arquivos/semiarido/arquivos/Relação_de_Municípios_Semiárido. pdf
https://www.gov.br/sudene/images/arquivo...
). As for the protected areas, dummy variables were generated to account for the presence of the municipal, state, and federal terrestrial Conservation Units (CU) located in Ceará’s territory (Brasil, 2020Brasil. Ministério do Meio Ambiente – MMA. (2020). Painel Unidades de Conservação. Retrieved in 2020, August 9, from https://app.powerbi.com/view?r=eyJrIjoiMDNmZTA5Y2ItNmFkMy00Njk2LWI4YjYtZDJlNzFkOGM5NWQ4IiwidCI6IjJiMjY2ZmE5LTNmOTMtNGJiMS05ODMwLTYzNDY3NTJmMDNlNCIsImMiOjF9
https://app.powerbi.com/view?r=eyJrIjoiM...
).

In addition, the municipal base has variables that derive from the crossing of two indicators from different sources. These variables indicate whole values as a proportion of the population or area of the municipality. Economic, tax, and electricity consumption information are expressed in per capita values, whereas cattle and population are expressed in terms of the municipal area.

## 3.2 Empirical Model

To estimate the EKC applied to the deforestation of forest remnants and non-forest natural areas in Ceará, we considered the model that includes the GDPpc variable in its squared and cubic forms, as follows:

D E F i t = β 0 + β 1 G D P p c i , t 1 + β 2 ( G D P p c ) i , t 1 2 + β 3 G D P p c i , t 1 3 + β k X i , t 1 + i t (1)

For which, DEFit is the annual deforestation for each of the i municipalities monitored by the “Aqui tem Mata?” program, where i=1,,61. The subscript t for the data set corresponds to the observed year t=2011,,2017, the GDPpci,t1 data set corresponds to the observed year denotes the level of economic activity of the i-th municipality in the previous year, and Xi,t1 corresponds to the set of additional explanatory variables lagged by one year (Table 2). The data compose a balanced panel with 427 observations.

From this, the EKC format is related to the sign and significance presented by the coefficients (β1,β2,β3). A sufficient condition for the EKC to present a linear format occurs when β1>0 or β1<0, while β2=β3=0. When β1>0an increase in GDPpc is linearly related to deforestation. For the “Inverted-U” format, it is sufficient that β1>0,β2<0and β3=0, for the “U” format, β10,β20and β3=0. Finally, in cases where β1>0,β2<0and β3>0 or β10,β20and β3<0, the curve takes the form of “N” and “N-inverted” respectively.

Note that, the dependent variable, the annual deforestation of the Atlantic Forest remnants and non-forest natural areas can be considered as a censored variable, as it is reasonable to assume that deforestation does not assume negative values, so we did not observe the variable of interest if it is below zero. In addition, the analysis of these data reveals that, of the 427 observations for deforestation, 347 are null observations (deforestation < 3 acres), that is, there is an excess of zeros and not the absence of information.

When disregarding such facts (censorship and excess of zeros) the traditional estimates of Ordinary Least Squares will be inconsistent since the assumptions of the classical linear model are violated. A model that applies well to these problems is the Tobit model, initially proposed by Tobin (1958)Tobin, J. (1958). Estimation of relationships for limited dependent variables. Econometrica, 26(1), 24-36., suitable when the dependent variable is censored or truncated, in addition to these cases, it is suitable when the dependent variable assumes an excessive number of zeros6 6 Do not confuse with models for inflated zero count data: Zero-Inflated Poisson (ZIP) or Zero-Inflated Negative Binomial (ZINB). (Calzolari et al., 2001Calzolari, G., Magazzini, L., & Mealli, F. (2001). Simulation-based estimation of Tobit model with random effects (pp. 349-369). Germany: MPRA.). Thus, the Tobit model can be expressed as follows:

Y i t = 0, i f Y i t * 0 Y i t * , i f Y i t * > 0 (2)

Where Yit*=Xit'β+it is observed only if strictly positive, Xit represents the vector of explanatory variables and the term itN0,σ2i.i.d and independent of Xit, with i=1,,N and t=1,,T. The probability density function when Y is observed is (Amemiya, 1985Amemiya, T. (1985). Advanced econometrics (pp. 536). Cambridge, MA: Harvard University Press.):

f ( Y i t | X i t ; θ ) = 0, i f Y i t < 0 Φ X i t ' β / σ ε , i f Y i t = 0 ϕ Y i t X i t ' β / σ ε , i f Y i t > 0 (3)

Where Φ is the cumulative distribution function and ϕ is the probability density function of the standard normal distribution. When considering the model for panel data, the error term it can be decomposed in (Calzolari et al., 2001Calzolari, G., Magazzini, L., & Mealli, F. (2001). Simulation-based estimation of Tobit model with random effects (pp. 349-369). Germany: MPRA.):

i t = α i + λ t + u i t (4)

Where αi are the individual effects (unobservable characteristics specific to unit i that are considered constant over time) interpreted as fixed parameters or as random variables, whereas λt is the effect of time (unobservable characteristics in period t, constant for all units cross-section in the sample) and uit is a random term that varies over time and with individuals, in addition uit is not correlated over time (Calzolari et al., 2001Calzolari, G., Magazzini, L., & Mealli, F. (2001). Simulation-based estimation of Tobit model with random effects (pp. 349-369). Germany: MPRA.).

Tobit regression models with panel data can take two forms, fixed effects or random effects. According to Cameron & Trivedi (2005)Cameron, A. C., & Trivedi, P. K. (2005). Microeconometrics: methods and applications (pp. 1058). New York: Cambridge University Press., in smaller panels (T < 8) the fixed effects estimator is not consistent, so the random-effects model is the most suitable:

Y i t * = α i + X i t ' β + u i t (5)

Where, αiN0,σα2i.i.d, uitN0,σu2i.i.d, withuit independent of αi, it is assumed that from the expression 4 the term λt=0t. The Equation 5 can be written as (Calzolari et al., 2001Calzolari, G., Magazzini, L., & Mealli, F. (2001). Simulation-based estimation of Tobit model with random effects (pp. 349-369). Germany: MPRA.):

Y i t * = σ α α i + X i t ' β + σ u u i t (6)

Where αiN0,1i.i.d independent ofuit, withuitN0,1i.i.d. Due to the individual effect, the observations on the dependent variable for each individual i are correlated. However, linked to the individual effect αi, the conditional joint density function can be written as (Gourieroux & Monfort, 1993Gourieroux, C., & Monfort, A. (1993). Simulation-based inference: a survey with special reference to panel data models. Journal of Econometrics, 59(1-2), 5-33.):

f Y i | X i , α i ; θ = t : Y i t > 0 1 σ u ϕ Y i t X i t ' β σ α α i σ u * t : Y i t = 0 Φ X i t ' β σ α α i σ u (7)

Φ and ϕ are the cumulative distribution function and the probability density function with the distribution N0,1. Since the individual effects αi are not observable, the equation 6 cannot be used in the inferences. To obtain maximum unconditional likelihood, it is necessary to integrate the individual effect αi:

f Y i | X i ; θ = f Y i | X i , α i ; θ d P α α (8)

A satisfactory solution for the integral 7 is the procedure via numerical integration. Alternatively, the integral 7 can be approximated using replicated simulations, thus obtaining a simulated likelihood function to be maximized to obtain a simulated maximum likelihood estimator (Calzolari et al., 2001Calzolari, G., Magazzini, L., & Mealli, F. (2001). Simulation-based estimation of Tobit model with random effects (pp. 349-369). Germany: MPRA.). Finally, as additional analyzes to the Tobit regression with random effects, there will be tested EKC with specifications of the GDPpc variable in its squared and cubic form for Tobit models with stacked or pooled Tobit data (base models).

# 4. Results and Discussion

The results are presented and discussed in three subsections that concentrate the analysis of the descriptive results, followed by the remaining and deforested areas and, finally, the econometric results are presented from the EKC specifications with GDPpc term in the quadratic and cubic forms, estimated by pooled Tobit and Tobit regressions with random effects.

## 4.1 Descriptive Data Analysis

Table 3 presents some descriptive statistics for the data set. In general, all variables showed a positive percentage variation in the range from 2011 to 2017, except for cattle farming, the percentage of jobs in the primary sector of the economy concerning the total of formal jobs and precipitation.

Table 3
Descriptive Statistics, 2011-2017.

The analysis of deforestation reveals an annual average of 6.93 acres/ municipality, with a maximum deforested extension of 375 acres in the municipality of Trairi (west coast of the state of Ceará). Of these observations, a large part (81%) refers to deforestation below 3 acres, so there are 347 null observations. Despite this fact, deforestation in these regions may be occurring due to the “ant effect”, deforestation small enough that the satellite does not capture its presence.

Regarding the economic information, the GDPpc stands out, with an average of R$10,000, with a minimum of R$ 2,700 for the city of Tururu and a maximum of R$71,000 in São Gonçalo do Amarante (SGA). For GAVpcagro the average was R$ 1.1 thousand, with a minimum registration in the city of Fortaleza and maximum value in Missão Velha. VEPpc, on the other hand, has an average of R$6.87, with a maximum per capita value in Frecheirinha (R$ 122.48).

Concerning POPD, the average is 263 inhabitants per km2, with the maximum density in the city of Fortaleza (capital of the state of Ceará) and minimum in the municipality of Granja. Cattle farming has an average of 14.60 heads/km2, with minimum and maximum records in the mountain towns of Guaramiranga, with 2.50 heads/km2, and Maranguape, with 34.41 heads/km2, located in the Metropolitan Region of Fortaleza (RMF).

In the descending part of the curve, R$23 thousand ≤ GDPpc ≤ R$ 58 thousand, there are four municipalities - Fortaleza, Aquiraz, Eusébio, and Maracanaú (the last three in the RMF), which account for just over 11% of deforestation. After the local minimum point (GDPpc ≥ R$58 thousand), in the ascending part of the curve, the municipality of São Gonçalo do Amarante is found, an important economic region in Ceará for contemplating the CIPP and also for having port and steel facilities, thermoelectric, wind farms and the Export Processing Zone (ZPE). From the EKC “N” shape, it can be inferred that the continuous increase in municipal economic activity, as measured by GDPpc, positively affects deforestation. And, as it increased again, it cannot be said that deforestation levels reached an absolute maximum in the region studied. The result is similar to that of Colusso et al. (2012)Colusso, M. V. S., Parré, J. L., & Almeida, E. S. (2012). Degradação ambiental e crescimento econômico: a curva de Kuznets ambiental para o cerrado. Revista de Economia e Agronegócio, 10, 335-357. who observed an EKC in “N” format for the relationship between economic growth and environmental degradation in the Cerrado biome. However, the different relationships between deforestation and pc income, established by the “Inverted-U” format (Ferreira & Coelho, 2015Ferreira, M. D. P., & Coelho, A. B. (2015). Desmatamento recente nos estados da Amazônia legal: uma análise da contribuição dos preços agrícolas e das políticas governamentais. Revista de Economia e Sociologia Rural, 53(1), 93-108.; Barros & Stege, 2019Barros, P. H. B., & Stege, A. L. (2019). Deforestation and human development in the Brazilian agricultural frontier: an environmental Kuznets curve for MATOPIBA. Revista Brasileira de Estudos Regionais e Urbanos, 13(2), 161-182.) and “Inverted-N” (Oliveira et al., 2011Oliveira, R. C., Almeida, E., Freguclia, R. S., & Barreto, R. C. S. (2011). Desmatamento e crescimento econômico no Brasil: uma análise da curva de Kuznets ambiental para a Amazônia Legal. Revista de Economia e Sociologia Rural, 49(3), 709-739.; Rodrigues et al., 2016Rodrigues, L. A., Cunha, D. A., Brito, L. M., & Pires, M. V. (2016). Pobreza, crescimento econômico e degradação ambiental no meio urbano brasileiro. Revista Iberoamericana de Economía Ecológica, 26(1), 11-24.) lead to the conclusion that the “Inverted-U” hypothesis for deforestation cannot be consolidated in the literature as a stylized fact. Finally, regarding the EKC hypothesis, we can state that the “Inverted-U” format is only verified in models with a quadratic specification of GDPpc. By incorporating the cubic term of GDPpc, EKC is configured in the form of “N”, breaking the paradigm that economic activity alone generates automatic protection for the environment, vis a vis sustainable development. On the contrary, the evidence found by this research points to the existence of a cyclical movement in the EKC in the remaining forest areas of the Atlantic Forest and non-forest natural areas in Ceará. # 5. Conclusion This research investigated the hypothesis of “U-inverted” for the EKC between GDPpc and annual deforestation of the forest and non-forest remnants of the Atlantic Forest in a sample of 61 municipalities from Ceará, monitored by the program “Aqui tem Mata?” from 2011 to 2017. The causes of deforestation are conditioned to several activities, especially anthropic ones. To contemplate these relationships, the characteristics associated with agriculture, consumption, demography, economic and tax information, the labor market, and geoenvironmental aspects are used. Understanding these relationships can assist in understanding the region's sustainable development process, therefore slowing deforestation activities in Atlantic Forest domains in Ceará. The relationship between deforestation and economic activity is verified in the form of “N”. In the ascending phase of the EKC (GDPpc ≤ R$ 23 thousand), deforestation is increasing for low levels of GDPpc. On the other hand, in the region of R$23 thousand to R$ 58 thousand of GDPpc, deforestation decreases as GDPpc rises, thus increasing again for GDPpc values above R$58 thousand. Another relevant aspect is that 91.8% of the municipalities are in the first rising phase of the curve since the average GDPpc of the State is R$ 10,000, which suggests that in the short term, most of these municipalities will not reach the intermediate range of the EKC (R$23 thousand ≤ GDPpc ≤ R$ 58 thousand), where dismantling decreases with the increase in economic activity. The EKC analyzes indicate that the GDPpc level is a crucial factor for deforestation.

Regarding the demographic pressure exerted by population density and by cattle farming, it appears that they reduce deforestation. However, it is worth mentioning that cattle ranching is not so expressive in Ceará and that the expansion of pasture areas can occur in the Caatinga domains, having no direct relationship with the Atlantic Forest and its natural non-forest forms. For population density, the effect can be associated with the preservation of forest fragments in urban and neighboring areas, due to the low availability of green areas in the region.

Given these results, the EKC’s “N” shape reveals a municipal behavior, in which economic development would not be aligned with the sustainable development of forest resources, since only four cities would be in the middle part of EKC, whereas, others 56 cities would still be in the first phase of the curve, with an intense relationship between economic activity and deforested area. Therefore, we concluded that the increase in GDPpc alone is not a factor that promotes an improvement in environmental quality.

The development of new technologies, as well as the unilateral efforts of the municipalities, enable to preserve the forest remnants and non-forest natural areas of the Atlantic Forest, thus reducing deforestation in the region, especially illegal deforestation. It is worth mentioning that the monitoring of these areas can assist in the design of public policies aimed at the sustainable use of natural resources and reduction of deforestation levels, as forests and associated ecosystems offer different environmental services and also, the protection of these wild habitats can prevent the transmission of zoonoses, due to the more direct interaction between humans and animals promoted by environmental suppression.

# Appendix A   Correlation matrix between explanatory variables.

• 1
Lowlands, natural altitude fields, vegetation refuges, dunes, herbaceous restinga, apicum, wetland and humid field.
• 2
Current Brazilian agricultural frontier, the MATOPIBA covers the Cerrado biome of the states of Maranhão, Tocantins, Piauí and Bahia, and accounts for a large part of the national production of grains and fibers (Empresa Brasileira de Agropecuária, 2020Empresa Brasileira de Agropecuária – EMBRAPA. (2020). Sobre o MATOPIBA. Retrieved in 2020, June 30, from https://www.embrapa.br/tema-matopiba/sobre-o-tema
https://www.embrapa.br/tema-matopiba/sob...
).
• 3
Almeida & Lobato (2019)Almeida, M. G., & Lobato, T. C. (2019). A curva de Kuznets ambiental para a região norte do Brasil entre os anos de 2002 a 2015. Economia & Região, 7(1), 7-25., Biage & Almeida (2015)Biage, M., & Almeida, H. J. F. (2015). Desenvolvimento e impacto ambiental: uma análise da curva ambiental de Kuznets. Pesquisa e Planejamento Economico, 45(3), 505-556. corroborate that the increase in economic activity increases deforestation via the agricultural sector with: burning in agriculture, extraction of forest resources, and pasture areas, which reduce CO2 absorption, increasing GHG emissions.
• 4
Water: biochemical oxygen demand (1990 to 2014), beach water quality (1992 to 2012). Atmosphere: anthropogenic GHG emissions (1990 to 2014), industrial consumption of ozone-depleting substances (1992 to 2013). Biodiversity (1992 to 2013): protected land areas, marine protection areas. Sanitation (1992 to 2011): access of the population to drinking water, sewerage and to the domestic garbage collection service. Land: deforestation of the Amazônia Legal (1990 to 2014), use of fertilizer (1992 to 2013), land use (1990 to 2011).
• 5
During the preparation of this survey, the municipal GDP for the year 2018 had not been released, so there is a limitation on the use of data on deforestation for the year 2017.
• 6
Do not confuse with models for inflated zero count data: Zero-Inflated Poisson (ZIP) or Zero-Inflated Negative Binomial (ZINB).
• How to cite: Sousa, W. L., Irffi, G., & Asevedo, M. D. G. (2022). Deforestation of the Atlantic Forest in the state of Ceará: analysis of the Environmental Kuznets curve from panel data, 2011 to 2017. Revista de Economia e Sociologia Rural, 60(1), e229884. https://doi.org/10.1590/1806-9479.2021.229884
• JEL Classification: C34, Q50, Q56

# 6. References

• Almeida, M. G., & Lobato, T. C. (2019). A curva de Kuznets ambiental para a região norte do Brasil entre os anos de 2002 a 2015. Economia & Região, 7(1), 7-25.
• Amemiya, T. (1985). Advanced econometrics (pp. 536). Cambridge, MA: Harvard University Press.
• Barros, P. H. B., & Stege, A. L. (2019). Deforestation and human development in the Brazilian agricultural frontier: an environmental Kuznets curve for MATOPIBA. Revista Brasileira de Estudos Regionais e Urbanos, 13(2), 161-182.
• Biage, M., & Almeida, H. J. F. (2015). Desenvolvimento e impacto ambiental: uma análise da curva ambiental de Kuznets. Pesquisa e Planejamento Economico, 45(3), 505-556.
• Brasil. Ministério da Integração – MI. (2018). Nova delimitação Semiárido Retrieved in 2020, August 9, from https://www.gov.br/sudene/images/arquivos/semiarido/arquivos/Relação_de_Municípios_Semiárido. pdf
» https://www.gov.br/sudene/images/arquivos/semiarido/arquivos/Relação_de_Municípios_Semiárido. pdf
• Brasil. Ministério do Meio Ambiente – MMA. (2020). Painel Unidades de Conservação Retrieved in 2020, August 9, from https://app.powerbi.com/view?r=eyJrIjoiMDNmZTA5Y2ItNmFkMy00Njk2LWI4YjYtZDJlNzFkOGM5NWQ4IiwidCI6IjJiMjY2ZmE5LTNmOTMtNGJiMS05ODMwLTYzNDY3NTJmMDNlNCIsImMiOjF9
» https://app.powerbi.com/view?r=eyJrIjoiMDNmZTA5Y2ItNmFkMy00Njk2LWI4YjYtZDJlNzFkOGM5NWQ4IiwidCI6IjJiMjY2ZmE5LTNmOTMtNGJiMS05ODMwLTYzNDY3NTJmMDNlNCIsImMiOjF9
• Brasil. Ministério do Trabalho – MTB. (2017). A Relação Anual de Informações Sociais - RAIS Retrieved in 2020, August 9, from http://www.rais.gov.br/sitio/index.jsf
» http://www.rais.gov.br/sitio/index.jsf
• Calzolari, G., Magazzini, L., & Mealli, F. (2001). Simulation-based estimation of Tobit model with random effects (pp. 349-369). Germany: MPRA.
• Cameron, A. C., & Trivedi, P. K. (2005). Microeconometrics: methods and applications (pp. 1058). New York: Cambridge University Press.
• Carvalho, T. S., & Almeida, E. S. (2010). A hipótese da curva de Kuznets ambiental global: uma perspectiva econométrico-espacial. Estudos Econômicos, 40(3), 587-615.
• Castelo, T. B., Adami, M., Almeida, C., & Almeida, O. T. (2018). Governos e mudanças nas políticas de combate ao desmatamento na Amazônia. Revista Iberoamericana de Economía Ecológica, 28(1), 125-148.
• Colusso, M. V. S., Parré, J. L., & Almeida, E. S. (2012). Degradação ambiental e crescimento econômico: a curva de Kuznets ambiental para o cerrado. Revista de Economia e Agronegócio, 10, 335-357.
• Cropper, M., & Griffiths, C. (1994). The Interaction of Population Growth and Environmental Quality. The American Economic Review, 84(2), 250-254.
• Delazeri, L. M. M. (2016). Determinantes do desmatamento nos municípios do arco verde - Amazônia Legal: uma abordagem econométrica. Economia Ensaios, 30(2), 11-34.
• Empresa Brasileira de Agropecuária – EMBRAPA. (2020). Sobre o MATOPIBA Retrieved in 2020, June 30, from https://www.embrapa.br/tema-matopiba/sobre-o-tema
» https://www.embrapa.br/tema-matopiba/sobre-o-tema
• Ente Nazionale per L’energia Elletrica – ENEL. (2017). Companhia Energética do Estado do Ceará – COELCE Retrieved in 2020, August 9, from https://www.enel.com.br/pt-ceara.html
» https://www.enel.com.br/pt-ceara.html
• Ferreira, M. D. P., & Coelho, A. B. (2015). Desmatamento recente nos estados da Amazônia legal: uma análise da contribuição dos preços agrícolas e das políticas governamentais. Revista de Economia e Sociologia Rural, 53(1), 93-108.
• Fundação Cearense de Meteorologia e Recursos Hídricos – FUNCEME. (2020). Portal Hidrológico do Ceará Retrieved in 2020, August 9, from http://www.hidro.ce.gov.br/municipios/chuvas-diarias
» http://www.hidro.ce.gov.br/municipios/chuvas-diarias
• Fundação SOS Mata Atlântica, & Instituto Nacional de Pesquisas Espaciais – INPE. (2020a). Atlas dos remanescentes florestais da Mata Atlântica: período 2018-2019 (pp. 61). São Paulo: INPE.
• Fundação SOS Mata Atlântica, & Instituto Nacional de Pesquisas Espaciais – INPE. (2020b). Aqui tem Mata? Retrieved in 2020, August 9, from https://aquitemmata.org.br/#/
» https://aquitemmata.org.br/#/
• Gourieroux, C., & Monfort, A. (1993). Simulation-based inference: a survey with special reference to panel data models. Journal of Econometrics, 59(1-2), 5-33.
• Grossman, G., & Krueger, A. (1995). Economic growth and the environment. The Quarterly Journal of Economics, 110(2), 353-377.
• Instituto Brasileiro de Geografia e Estatística – IBGE. (2017a). Produto Interno Bruto dos Municípios Retrieved in 2020, August 9, from https://www.ibge.gov.br/estatisticas/economicas/contas-nacionais/9088-produto-interno-bruto-dos-municipios.html?t=downloads
• Instituto Brasileiro de Geografia e Estatística – IBGE. (2017b). Sistema IBGE de Recuperação Automática – SIDRA Retrieved in 2020, August 09, from https://sidra.ibge.gov.br/tabela/5938
» https://sidra.ibge.gov.br/tabela/5938
• Instituto Brasileiro de Geografia e Estatística – IBGE. (2017c). Produção da Extração Vegetal e da Silvicultura – PEVS Retrieved in 2020, August 09, from https://www.ibge.gov.br/estatisticas/economicas/agricultura-e-pecuaria/9105-producao-da-extracao-vegetal-e-da-silvicultura.html?=&t=o-que-e
» https://www.ibge.gov.br/estatisticas/economicas/agricultura-e-pecuaria/9105-producao-da-extracao-vegetal-e-da-silvicultura.html?=&t=o-que-e
• Instituto Brasileiro de Geografia e Estatística – IBGE. (2017d). Estimativas da População Retrieved in 2020, August 09, from https://www.ibge.gov.br/estatisticas/sociais/populacao/9103-estimativas-de-popul?=&t=downloads
• Instituto Brasileiro de Geografia e Estatística – IBGE. (2020). Pesquisa da Pecuária Municipal - PPM Retrieved in 2020, August 09, from https://www.ibge.gov.br/estatisticas/economicas/9107-producao-da-pecuaria-municipal.html?=&t=downloads
• Oliveira, R. C., Almeida, E., Freguclia, R. S., & Barreto, R. C. S. (2011). Desmatamento e crescimento econômico no Brasil: uma análise da curva de Kuznets ambiental para a Amazônia Legal. Revista de Economia e Sociologia Rural, 49(3), 709-739.
• Rodrigues, L. A., Cunha, D. A., Brito, L. M., & Pires, M. V. (2016). Pobreza, crescimento econômico e degradação ambiental no meio urbano brasileiro. Revista Iberoamericana de Economía Ecológica, 26(1), 11-24.
• Sistema de Informações Contábeis e Fiscais do Setor Público Brasileiro – SICONFI. (2017). Contas anuais Retrieved in 2020, August 9, from http://www.tesouro.fazenda.gov.br/contas-anuais
» http://www.tesouro.fazenda.gov.br/contas-anuais
• Soares, L. R., & Almeida, L. T. (2018). Desacoplamento de impactos ambientais no Brasil. Revista Iberoamericana de Economía Ecológica, 28(2), 21-43.
• Teixeira, R. F. A. P., Bertella, M. A., & Almeida, L. T. (2012). Curva de Kuznets ambiental para o estado do mato grosso. Análise Econômica, 30(57), 313-337.
• Tobin, J. (1958). Estimation of relationships for limited dependent variables. Econometrica, 26(1), 24-36.

# Publication Dates

• Publication in this collection
14 June 2021
• Date of issue
2022