The impact of deforestation, urbanization, public investments, and agriculture on human welfare in the Brazilian Amazonia

Land Use Policy 65 (2017) 135–142

Contents lists available at ScienceDirect

Land Use Policy journal homepage: www.elsevier.com/locate/landusepol

The impact of deforestation, urbanization, public investments, and agriculture on human welfare in the Brazilian Amazonia

MARK

⁎

José Maria Cardoso da Silvaa, , Shivangi Prasada, José Alexandre Felizola Diniz-Filhob a b

University of Miami, Department of Geography, 1300 Campo Sano Avenue, P.O. Box 248067, 33124-4401, Miami, FL, USA Universidade Federal de Goiás, Instituto de Ciências Biológicas, Departamento de Ecologia, 74690-970, Goiânia, GO, Brazil

A R T I C L E I N F O

A B S T R A C T

Keywords: Sustainability Human welfare Local development Development convergence Brazil Amazonia

The relationship between human welfare and deforestation in the Brazilian Amazonia has traditionally been thought to follow a boom-and-bust pattern. According to this pattern, forest clearing triggers rapid increases in human welfare levels (“the boom”) due to short-term economic gains; these levels then drop to below national or regional averages (“the bust”) after the forest stocks have declined, thus causing the local populations to become deprived of ecosystem services. However, recent studies have questioned the validity of this boom-and-bust pattern. In this paper, we use panel data and simultaneous autoregressive models to evaluate the effects of deforestation, urbanization, public investments, agriculture, and state policies on temporal changes in human welfare that occurred across multiple municipalities in the Brazilian Amazonia from 2005 to 2012, a period during which governments implemented a set of strategies aimed at controlling deforestation across the region. We found that: (a) signals of a boom-and-bust pattern are weak at the regional level, and therefore this pattern cannot be generalized across the entire region; (b) human welfare is increasing more rapidly in low-development municipalities than in high-development cities, and all municipalities are converging on at least one regional average rather than on a national average; (c) urbanization does not lead to positive changes in human welfare, which indicates that the infrastructure available in regional urban centers is limited; (d) public investments are negatively associated with human welfare growth, thus signifying that if public investments are not used to leverage the potential of other sectors of the local economy, human welfare will not improve; (e) agriculture is negatively associated with positive changes in human welfare at the local level, possibly due to the dominance of cattle-ranching as the predominant economic activity of this sector; and (f) state-level policies matter, and future analyses of regional trends in the realm of development and conservation across this region should take such policies into account. Finally, we suggest that although human welfare and deforestation retain a weak statistical relationship, we cannot contend that they have been fully decoupled. Forest loss across the region is still pervasive, and institutions are too weak to sustain the transition from a frontier development model to a conservation-centered model.

1. Introduction In tropical forest regions, traditional development usually follows the frontier model, in which forests are replaced by other types of land better suited to the production of quick economic gains. Such a model does not embrace long-term concern about environmental sustainability (Becker, 2001). In places where financial resources generated by the depletion of forest stocks are reinvested into the local community, average human welfare is likely to improve. Over time, a favorable standard of living can lead to better environmental regulations and the advent of local organizations. These factors can, in turn, lead to forest transition (i.e., a reversal from net forest loss to net forest gain)

⁎

Corresponding author. E-mail address: [email protected] (J.M.C.d. Silva).

http://dx.doi.org/10.1016/j.landusepol.2017.04.003 Received 13 July 2016; Received in revised form 28 March 2017; Accepted 1 April 2017 0264-8377/ © 2017 Elsevier Ltd. All rights reserved.

(Mather, 1992; Rudel et al., 2005). However, if the revenues generated by forest use are exported to other places, and if local organizations are not capable of adopting actions that sustain a relatively high growth rate in human welfare levels, a different pattern can occur. Instead of forest transition, such places may exhibit a boom-and-bust pattern. This pattern is found when the short-term gains caused by forest clearing trigger a rapid growth rate in average human welfare (“the boom”) which then drops to below national or regional averages (“the bust”) after the forest stocks have declined, thus causing populations to become deprived of the ecosystem services that once sustained their economic activities (Schneider et al., 2002; Rodrigues et al., 2009). An alternative to the frontier model is the conservation-centered model, in

Land Use Policy 65 (2017) 135–142

J.M.C.d. Silva et al.

that the sector could have a positive social impact on local societies. However, since the 1970s, agricultural expansion in the Brazilian Amazonia has caused several social-ecological issues by exacerbating land use conflicts, undermining the rights and lifestyles of traditional people, and facilitating the spatial diffusion of homicidal violence (Becker, 2004; Arima et al., 2005; Hecht, 2011; Silva, 2015). In addition to social problems, monocultures and cattle-ranching were and continue to be the major drivers of massive regional deforestation (Gibbs et al., 2015; Hansen et al., 2013), thereby eroding important ecosystem services that are relevant to both local and global societies (Fearnside, 1997; Clement and Higuchi, 2006; Silva, 2015). Consequently, whether monocultures and cattle ranching make a genuine contribution to positive changes in human welfare across the Brazilian Amazonia remains a controversial issue (Prates and Bacha, 2011). Most of the recent discussion about human welfare across the Brazilian Amazonia is focused on the policies and programs designed by the national government. Almost no emphasis is placed on statelevel initiatives. However, Brazil is a federative republic wherein states have autonomy—and their own resources—to set policies and programs that can converge on or diverge from the national agenda. In the last decades, states have become strong protagonists of the region’s socioeconomic development, designing and leading the implementation of innovative public policies (Silva et al., 2005; Garda et al., 2010). However, the strategies adopted by the various states within the region are different. For instance, while some of the states in the Brazilian Amazonia (such as Amazonas, Amapá, Acre, and Pará) have embraced a conservation-centered development plan in at least a portion of their territories, other states (such as Mato Grosso, Rondônia, Roraima, Maranhão, and Tocantins) continue to base their development strategies on the traditional frontier model (Silva et al., 2005; Garda et al., 2010). These differences among the many state-level development strategies are predicted to lead to disparities between local development trajectories. In this study, we evaluated the effects of deforestation, urbanization, public investments, agriculture, and state policies on temporal changes in human welfare across various municipalities in the Brazilian Amazonia by using panel data from 2005 to 2012, the period immediately following the implementation of the PPCDAm. To this end, we assessed the following hypotheses: (a) signals of a boom-andbust pattern are weak across the region; (b) human welfare in the region is converging on a rising national average; (c) increased urbanization contributes positively to temporal changes in human welfare; (d) increased public investments contribute positively to temporal changes in human welfare; and (e) agriculture contributes positively to temporal changes in human welfare. Additionally, because the Brazilian Amazonia is a heterogeneous region composed of nine states, all with different policy priorities and socioeconomic contexts, we evaluated the hypothesis that the set of relationships described above also holds at the state level. Finally, we applied our results to a discussion of whether—as suggested by Caviglia-Harris et al. (2016)—the socio-economic development that occurred across the region during our study period became fully decoupled from deforestation.

which societies use knowledge and technology to design sustainable territories wherein most natural ecosystems are protected or wisely used and where human welfare improves as a consequence of the development of local economies that are diversified, efficient, inclusive, and resilient (Becker, 2004; Vieira et al., 2005; Silva, 2015; Nobre et al., 2016). The Brazilian Amazonia is considered to be a textbook example of the modern frontier development model. Since the 1960s, the expansion of roads, dams, and large mineral projects has led to an intense process of regional occupation that has already claimed roughly 20% of the original forests and caused recurrent social conflicts (Becker, 2004; Hecht, 2011; Silva, 2015; Souza et al., 2015). From 1988–2015, 413,505 km2 of forests in the region were replaced by other types of land use (INPE, 2015). In 2004, deforestation peaked at 27,772 km2, which led the Brazilian government to design a long-term plan to control deforestation and move the region away from the traditional frontier model to a more conservation-centered development plan (Hecht, 2011). The Action Plan for Prevention and Control of Deforestation in the Legal Amazonia (PPCDAm) combined a set of initiatives, including those focused on the expansion of protected areas, the recognition of indigenous lands, the increased enforcement of existing environmental legislation, the development of a new forest monitoring system led by the National Institute for Spatial Research (INPE), the creation of incentives for forest production, and the reduction of subsides and credits for economic activities that sustain illegal deforestation (Hecht, 2011; Assunção et al., 2015; Rajão and Georgiadou, 2014). Aligned to PPCDAm there were various private sector initiatives such as the soybean and beef moratoria (Gibbs et al., 2015, 2016). To date, the PPCDAm has been successful, and resulted in deforestation decline to its lowest historical rate of 4571 km2 in 2012 (INPE, 2015). However, because forest conservation is a historical and political process, only a substantial increase in human welfare across the region will ensure the long-term maintenance of low deforestation rates and accelerate the transition to a conservation-centered development model (Vieira et al., 2005; Dias et al., 2016; Aguiar et al., 2016). The relationship between human welfare and deforestation in the Brazilian Amazonia remains a controversial issue. Schneider et al. (2002) proposed that this relationship follows the boom-and-bust pattern rather than the forest transition pattern. Rodrigues et al. (2009) and Celentano et al. (2012) evaluated and supported this hypothesis by using cross-sectional data from the year 2000. However, recent efforts to analyze panel data (from 1990 to 2010) conducted by Caviglia-Harris et al. (2016) and Weinhold et al. (2015) have questioned these findings. Weinhold et al. (2015) suggested that the results of the cross-sectional data analysis are merely artifacts of spatial correlation and that municipalities with different levels of forest cover have enjoyed equal increases in human welfare over a decade, with no evidence of a boom-and-bust pattern. Caviglia-Harris et al. (2016) found that a weak but significant boom-and-bust pattern was identified only when human welfare rates during the first year of the study period were included in the regression model. Furthermore, Caviglia-Harris et al. (2016) suggested the following: (a) that human welfare in the region has become decoupled from deforestation, (b) that human welfare rates are converging on rising national averages, and (c) that this convergence is the result of an increase in public investments and the rapid urbanization process that is occurring throughout the region. In the search for factors that can explain the geographic patterns of temporal changes in human welfare across the Brazilian Amazonia, two factors have not yet been assessed: agriculture and state-level policies. The most recent wave of human occupation in the Brazilian Amazonia was based on the assumption that the promotion of monocultures and cattle ranching as a regional development strategy would lead to substantial gains in local human welfare (Becker, 2004). Supporting this assumption is the fact that, from a purely economic viewpoint, agriculture is—among the major economic sectors of Brazil—the segment with the smallest Gini coefficient (IBGE, 2014). This indicates

2. Methods 2.1. Study area We delimited the Brazilian Amazonia according to the boundaries of the Amazonia Biome as defined by the Brazilian Institute of Geography and Statistics (IBGE, 2004). The IBGE’s proposal follows the boundaries laid out in the original extension of the tropical rainforests of northern Brazil. This region covers an area of 4.3 million km2 (Fig. 1) and has a population of 21.6 million people, 72% of whom live in urban areas. The Brazilian Amazonia includes 517 municipalities in the following nine Brazilian states: Amazonas, Acre, Rondônia, Roraima, Amapá, 136

Land Use Policy 65 (2017) 135–142

J.M.C.d. Silva et al.

Fig. 1. States and municipalities in the Brazilian Amazonia. Note that only a portion of the municipalities in Maranhão, Tocantins, and Mato Grosso are considered part of the Brazilian Amazonia.

of the local gross domestic product (GDP) that was generated in 2012 through public services (IBGE, 2015). This included the gross value that was added as a result of policies such as public management (including the salaries of public servants), public education, health, and social security (IBGE, 2008). This indicator of public investments can be considered reliable because most financial flows are documented by governments and are subject to scrutiny from multiple public and private organizations. To assess the agricultural sector’s contribution to human welfare, we used as an indicator the contribution of the agricultural sector (in percentages) to the local GDP in 2012 (IBGE, 2015). Use of this indicator has some limitations, and we applied it to our analyses while taking into consideration two important caveats. First, this indicator includes the gross economic value generated by four distinct economic activities: agriculture, cattle ranching, forestry, and fisheries (IBGE, 2008). Second, an unknown but possibly significant portion of the gross value of these four economic sectors is traded in informal markets and consequently is not captured in the official statistics.

Pará, Mato Grosso, Maranhão, and Tocantins. 2.2. Data We evaluated changes in human welfare, forest cover, urbanization, public investments, and agricultural output in the majority of the municipalities located across the Brazilian Amazonia from 2005 to 2012. We used the Index FIRJAN of Municipal Development (IFDM) as the indicator of human welfare at the local level. This index is generated annually for all municipalities in Brazil by using standard indicators produced by the Brazilian government. The IFDM is intended to represent overall local socioeconomic welfare by evaluating progress in three essential components of human wellbeing: jobs and income, education, and health. Within each component, five to six indicators that are produced annually by the Brazilian Ministries of Labor, Education, and Health are aggregated and weighted in order to construct a sub-index for each component, with values ranging from 0 to 1. Finally, these three sub-indices are aggregated using equal weights to compose the IFDM. The full methodology, which shows the variables used for all calculations, sources, and relative weights, is described by FIRJAN (2015). The cumulative deforestation rate for 2012 was obtained from the PRODES Project of the Brazilian Institute for Space Research (INPE, 2015). As a proxy for urbanization, we used figures representing the proportion of the urban population of each municipality, as found in the national population census of 2010 (IBGE, 2011). We evaluated public investments across municipalities by examining the percentage

2.3. Analyses To test our hypotheses, we used several forms of regression to model changes in human welfare from 2005 (T0) to 2012 (T1). For a few municipalities, information about human welfare was not available for 2005 or 2012; in these cases, we calculated the change by using the next year (2006 or 2007) for which the information was available. We divided the difference in human welfare between T1 and T0 by the 137

Land Use Policy 65 (2017) 135–142

J.M.C.d. Silva et al.

Table 1 Means, standard deviations (SD), and coefficients of variation (CV) for the Municipal Human Development Index (HDI) in both T0 (2005) and T1 (2012)a for 499 municipalities in the Brazilian Amazonia. The indices of all Brazilian municipalities were added for comparison. States

Acre Amazonas Amapá Maranhão Mato Grosso Pará Rondônia Roraima

Number of Municipalities

Human Development Index Mean (T0)

SD (T0)

CV (T0)

Mean (T1)

SD (T1)

CV (T1)

21 59 16 86 83 140 52 13 29

0.39 0.36 0.47 0.37 0.54 0.39 0.49 0.45 0.54

0.09 0.07 0.08 0.07 0.08 0.08 0.07 0.09 0.07

0.23 0.19 0.17 0.19 0.15 0.21 0.14 0.20 0.13

0.51 0.46 0.54 0.49 0.67 0.49 0.6 0.52 0.64

0.09 0.08 0.07 0.08 0.08 0.09 0.08 0.10 0.06

0.18 0.17 0.13 0.16 0.12 0.18 0.13 0.19 0.09

499 5565

0.43 0.51

0.10 0.18

0.23 0.35

0.54 0.63

0.11 0.12

0.2 0.2

Tocantins All region Brazil a

For some municipalities, T0 is 2006 or 2007. The Municipal Human Development Index was obtained from FIRJAN (2015).

b

eliminate the spatial structure of data, the OLS model was used to obtain the expected slopes under the null hypothesis (the absence of a relationship between the change in the human welfare index and its initial value). The regression models are useful to test whether municipalities with low human welfare rates are improving their conditions more rapidly than are municipalities with high human welfare scores. This concept is called β-convergence (Barro and Sala-i-Martin, 1995). However, to test whether human welfare in the Brazilian Amazonia is converging on a national or regional average, it is also necessary to evaluate the σconvergence. The simplest and most efficient method of evaluating such σ-convergence is to compare the coefficients of variation in the human welfare index across two time periods (Castro, 2004). We performed all analyses in the R-platform (R Development Core Team, 2015), and SAR was fitted using the spdep package (see Bivand et al., 2013).

number of years under study in order to take these differences in data collection into account. We excluded from our analyses 18 municipalities whose comparison periods (i.e., the difference between T1 and T0) were less than five years. We used cumulative deforestation, urbanization, public investments, and agriculture in T1 as predictors of the model. Because the boom-and-bust pattern predicts a quadratic relationship between cumulative deforestation and human welfare, we added both cumulative deforestation in T1 and the square of this variable to the model (Celentano et al., 2012; Weinhold et al., 2015; Caviglia-Harris et al., 2016). We included human welfare in T0 (log-transformed) as an explanatory variable to take into account the temporal dependency of the changes in human welfare variable (Caviglia-Harris et al., 2016). Finally, we also added the state in which each municipality is located as a factor in the model in an effort to consider potential variation in human welfare that results from policies at the state level. Because this effect was significant, we modeled changes in human welfare separately for each state as well. We began to model changes in human welfare by using an ordinary least squares (OLS) regression. However, a Moran’s I coefficient based on connections between municipalities situated less than 350 km apart revealed a small but significant spatial autocorrelation in model residuals (I = 0.075; P = 0.007 using 1000 randomizations), which can bias significance tests (see Legendre and Legendre, 1998). Thus, we applied a simultaneous autoregressive model (SAR) (Cressie, 1993; Haining, 2002; Fortin and Dale, 2005), thus incorporating the spatial connectivity of municipalities in order to take into account the autocorrelation in model residuals. The SAR model efficiently controlled for autocorrelation in model residuals (I = 0.024; P = 0.191 using 1000 randomizations). An adjusted pseudo-R2 was calculated for SAR by correlating observed and fitted values. A comparison of models that included the effects of states was completed using Akaike information criterion (AIC) (Burnham and Anderson, 2002). It is also important to note that because Caviglia-Harris et al. (2016) modeled changes in the human welfare index using the initial condition (i.e., the value of human welfare in T0), there is an intrinsic correlation between the explanatory and response variables (e.g., Jackson and Somers, 1991). Therefore, we tested the statistical significance of the effect of this variable using a null model. We randomized the human welfare index in T0, added a random variable with the mean and standard deviation of the observed shift in human welfare from T0 to T1, and then calculated the index. The regression model was then repeated 1000 times using randomized data in order to build a null distribution of coefficients to which the observed coefficient of changes in human welfare could be compared. Because randomization tends to

3. Results We collected information on 499 municipalities in the Brazilian Amazonia; these municipalities covered 4.28 million km2 and were home to 20 million people in 2012. In general, human welfare improved across all states at the regional level at a rate similar to the national pace (Table 1); however, municipalities in states other than Tocantins and Mato Grosso had an average human welfare score that was lower than the Brazilian average (Table 1). The coefficients of variation in the human welfare index during the study period declined at similar rates at both the regional and state levels, though these two rates were lower than Brazil’s average (Table 1). In 2012, average cumulative deforestation rates within municipalities ranged from 3.66% in Amapá to 70.55% in Tocantins (Table 2). States whose municipalities had greater levels of forest cover also demonstrated less variation in cumulative deforestation rates between municipalities, as indicated by the standard deviation. Furthermore, urbanization was quite consistent across the region’s municipalities; scores ranged from 0.40 in Roraima to 0.69 in Amapá, with a regional average of 0.54 that was lower than the national average (Table 2). The average contribution of public investments to the local GDP showed significant variation between states; contributions ranged from 22.50% in Mato Grosso to 67.06% in Roraima, with a regional average of 33.61% that was only slightly higher than the national average (Table 3). On average, however, municipalities in eight of the nine states had GDP levels that were more dependent on public services than the Brazilian average (Table 3). The agricultural output represented, on average, 25.11% of the GDP of Amazonian municipalities (Table 3), 138

Land Use Policy 65 (2017) 135–142

J.M.C.d. Silva et al.

Table 2 Means and standard deviations (SD) of cumulative deforestation (in 2012), urbanization (in 2010), percentage of the gross domestic product (GDP) generated through public services (in 2012), and percentage of the GDP generated through agriculture (in 2012) for 499 municipalities in the Brazilian Amazonia, summarized by state. Information on urbanization as well as the contributions of public services and agriculture to the GDP of all Brazilian municipalities was added for comparison. States

Number of Municipalities

Cumulative Deforestationa

Urbanizationb

Mean (T0)

Mean

SD (T0)

SD

Public Investmentsc

Agricultured

Mean

Mean

SD

SD

Acre Amazonas Amapá Maranhão Mato Grosso Pará Rondônia Roraima Tocantins

21 59 16 86 83 140 52 13 29

25.57 5.06 3.66 67.54 46.76 46.82 56.09 4.79 70.55

23.98 5.8 2.92 26.35 24.18 30.38 25.58 5.76 20.62

0.54 0.55 0.69 0.50 0.63 0.52 0.52 0.40 0.66

0.18 0.15 0.19 0.19 0.16 0.2 0.22 0.23 0.16

49.99 46.96 55.82 41.42 22.50 35.09 33.56 67.06 42.20

13.04 15.19 11.65 11.32 10.15 13.27 7.48 12.03 14.05

24.57 27.62 11.25 19.81 32.93 27.26 28.29 10.51 18.73

11.85 14.43 9.41 9.93 17.89 15.04 14.22 7.72 11.76

All region Brazil

499 5565

44.4

32.14

0.54 0.64

0.19 0.22

33.61 31.49

13.99 19.07

25.11 18.74

15.07 14.95

a

Cumulative deforestation rates were obtained from INPE (2015). Urbanization is the proportion of the population living in urban centers (IBGE, 2004). The percentage of the local gross domestic product generated through public services (IBGE, 2015). d The percentage of the local gross domestic product generated through agriculture (IBGE, 2015). b c

improving more rapidly than are municipalities with a high human welfare index. The hypothesis that urbanization is positively associated with human welfare across the region is not supported by our model (b = 1.442 – P > 0.05). We found that both public investments and agriculture were associated with changes in human welfare; however, against the proposed hypothesis, the effects of these two variables were negative rather than positive (Table 3). Again, though, public investments and agricultural output alone explain only a tiny percentage of the changes in human welfare across the region (a very small R2 of 3.56%). Our hypothesis that variation between states will be revealed following the application of our general model is supported (AIC = 32.3). Although adding the effects of states did not cause a strong shift in the overall coefficients of the explanatory variables, the relative importance of these different predictors of human welfare resulted in wide variation among Brazil’s nine states (Table 3). The models that were applied at the state level were much better than those

with less dependency in Roraima (10.51%) and more in Mato Grosso (32.93%). The hypothesis that changes in human welfare and cumulative deforestation follow a boom-and-bust pattern at the regional level is partially supported by our model, as we found a positive coefficient with cumulative deforestation and a negative coefficient with squared cumulative deforestation, both of which were significant at P < 0.05 according to the SAR model (Table 3). However, these two variables alone explain only a minor portion of the variation in changes in human welfare (2.6%). The explanatory power of the SAR model increased to 24.6% when the human welfare index in T0 and other variables were incorporated, and the coefficient was significant at P < 0.01 under a randomization model that took into account the intrinsic dependence of the response and explanatory variables (with a negative coefficient) (Table 3). Thus, changes in current human welfare across the region were strongly dependent on the initial conditions of T0; this supports the hypothesis that municipalities with a low human welfare index are

Table 3 Coefficients of each explanatory variable (i.e., the standard normal deviate of regression coefficients) on the changes in the Human Development Index between 2005 and 2012 obtained from Simultaneous Autoregressive (SAR) spatial regressions. The models are for the entire Brazilian Amazon, the Brazilian Amazon in combination with the effects of states as a dummy variable, and for each state separately. For each model, it is also included the adjusted coefficient of determination (R2). Values in bold indicate significant effects at P < 0.05. The statistical significance of Log (IDH T0) was obtained via a null model analysis with 1000 randomizations (see text for details). All Region

Region + States

ROa

AC

AM

RR

PA

AP

TO

MA

MT

−11.658 2.789

−13.294 2.979

−4.039 0.474

−1.825 0.466

−5.944 1.114

−7.264 6.919

−8.360 1.176

−2.090 2.122

−6.224 −2.322

−5.108 −0.729

−5.589

Deforestation T1c (quadratic effect)

−2.486

−2.641

−0.095

−0.309

−0.655

−5.300

−0.680

−2.181

2.377

0.484

Public Investmentsd

−6.065

−8.030

−3.449

−1.106

−3.324

−7.137

−4.279

0.265

−2.538

−3.399

1.442

0.851

−1.791

−0.861

1.156

−3.552

0.744

−0.586

−1.578

1.429

−4.331

−4.744

−3.350

−0.371

−2.206

−5.368

−4.472

−0.491

−3.310

−1.004

0.246

0.324

0.443

0.627

0.391

0.931

0.369

0.742

0.791

0.322

Explanatory Variable Log (Human Development T0)b Deforestation T1c

2.179 −2.236 −2.892 Urbanizatione

0.136 Agriculturef

−1.652 Adj_R2

0.344 n

499

499

52

21

59

13

140

16

29

86 82

a

States are: RO (Rondônia), AC (Acre), AM (Amazonas), RR (Roraima), PA (Pará), AP (Amapá), TO (Tocantins), MA (Maranhão), and MT (Mato Grosso). Data was obtained from the Municipal Human Development Index (FIRJAN, 2015). Cumulative deforestation rates were obtained from INPE (2015). d The percentage of the gross domestic product generated through public services (IBGE, 2015). e Urbanization was measured as the proportion of the population living in urban centers (IBGE, 2015). f The percentage of the gross domestic product generated through agriculture (IBGE, 2015). b c

139

Land Use Policy 65 (2017) 135–142

J.M.C.d. Silva et al.

that were applied regionally, with pseudo-R2 ranging from 0.32 in Maranhão to 0.93 in Roraima (note, however, that these R2 also incorporate the spatial effects of changes in human welfare and do not necessarily indicate the strong effects of explanatory variables). In all states, the log of human welfare in T0 was significant and was the most important explanatory variable. The effect of cumulative deforestation (including the quadratic effect) was significant only in the states of Roraima, Amapá, Maranhão, and Tocantins (Table 3). In Roraima, Amapá, and Maranhão, the results fit with the predictions of the boom-and-bust model, while in Tocantins, that trend was inverse; thus, the greatest changes in human welfare occurred in municipalities with either lower or higher deforestation rates (whereas according to the boom-and-bust pattern, changes in human welfare occur at intermediate levels of deforestation). Furthermore, urbanization had a significant and negative relationship with human welfare only in the state of Roraima. Public investments and agriculture demonstrated negative associations with changes in human welfare in all states, though they were significant in different state combinations (Table 3). Public investments, for instance, were significant in all states except Acre and Amapá. In contrast, agriculture was significant in five states (Roraima, Amazonas, Rondônia, Pará, and Tocantins).

human welfare between the municipalities has reduced over time, thus indicating σ-convergence. However, the rate of σ-convergence across the Brazilian Amazonia was lower than the national rate, which suggests that the region’s municipalities are not converging on a national average, as was suggested by Caviglia-Harris et al. (2016). In fact, because Brazil is large and multifaceted, it is possible that the country has multiple equilibrium points (or “convergence clubs”) with regard to human welfare rather than just one. Recent studies on development convergence in Brazil have recognized regional differences in development pathways at the local level and have indicated the existence of multiple “convergence clubs” within the country (review in Barberia and Biderman, 2010). Therefore, because most municipalities across the region still fall well below the national human welfare average, and because there are important differences among the various states, we suggest that the majority of Amazonian municipalities are possibly converging on at least one regional average rather than on a single rising national average. Caviglia-Harris et al. (2016) suggested that development convergence across the Brazilian Amazonia could be due to two factors: rapid urbanization and public investments. Rapid urbanization is an ongoing process in the region, as more and more rural people move to cities in search of better public services and more favorable job opportunities (Becker, 2001, 2004). In the Brazilian Amazonia, urbanization occurred at a more rapid pace than in the country as a whole, increasing from 37% in 1970 to 73% in 2010 (IBGE, 2011). Still, the average proportion of people living in urban centers across the region’s municipalities is lower than the percentage of the population living in the country’s municipalities. Currently, most of the regional population lives in less than 1% of the region’s land; this could lead to a reduction in deforestation if cities were not dependent on the commodities produced in adjacent rural territories. However, there is recent evidence that the most rapid urban growth in the region is occurring within cities that are located near rural areas that produce commodities (minerals or crops) and are connected to export corridors (Richards and VanWey, 2015). In this study, we found that urbanization is not significantly associated with positive changes in human welfare indicators at the regional level. At the state level, only Roraima had a significant relationship between urbanization and changes in human welfare, though this relationship was negative rather than positive. Urbanization can improve human welfare if urban centers have the infrastructure required to offer good and reliable public services to the population. However, this is not the case for most of the cities in the Brazilian Amazonia, wherein the rapid urban growth caused by intense rural–urban migration has not been followed by massive investments in infrastructure to accommodate the new demand for jobs and public services (Perz, 2000). As a result, cities across the region have urban infrastructure and public service quality levels that are several degrees below the standards found in other Brazilian regions (Santos et al., 2014). This leads to intense social conflicts that over time undermine genuine advances in local human welfare (Souza et al., 2015). The hypothesis that public investments would contribute positively to the improvement of human welfare (Caviglia-Harris et al., 2016) is not supported by our analysis. In fact, we found a negative relationship between the two variables, thus indicating that municipalities with economies that are strongly dependent on public investments are those that presented low human welfare growth rates. At the state level, the negative relationship between human welfare growth and public investments is significant in all states except Acre and Amapá. The negative relationship between positive changes in human welfare and public investments suggests that greater levels of public investment will not necessarily sustain positive changes in human welfare at the local level if these investments are not used efficiently to foster sustainable economic activities that increase the contributions of other sectors of the economy. The hypothesis that agriculture would contribute positively to the improvement of human welfare is not supported by our regional model,

4. Discussion The Brazilian Amazonia is a large and complex region that faces the challenge of reconciling economic development with the conservation of its natural resources for the sake of local, national, and global societies (Fearnside, 1997; Clement and Higuchi, 2006; Silva, 2015). The Brazilian government’s decision in 2004 to replace the traditional frontier development model with a new model based on the sustainable use of the region’s natural resources has already reduced deforestation rates and improved human welfare across the region. Our results provide a more nuanced perspective on the major factors that drove changes in human welfare in various local municipalities across the region during the seven years (2005–2012) following the implementation of large-scale governmental efforts to shift the regional development trajectory. Thus, this study contributes to a better understanding of the frontier dynamics at play within the world’s largest tropical forest. We found that signals of a boom-and-bust pattern at the regional level were significant but weak because the variation in human welfare that was explained by cumulative deforestation or squared cumulative deforestation (as well as by the other analyzed variables) was very low. This finding is similar to the patterns described by Caviglia-Harris et al. (2016), who analyzed panel data from 1990 to 2010. However, because these authors did not take into account the temporal and spatial dependence of the variables used in their model, their R2 were higher than those that we reported. Moreover, the weak regional signals of a boom-and-bust pattern vanished at the state level. In six of the nine states, there was no relationship whatsoever between changes in human welfare and cumulative deforestation or squared cumulative deforestation between 2005 and 2012. Of the three states that exhibited indications of a boom-and-bust pattern, one is located in the old economic frontier of the southern Brazilian Amazonia (Mato Grosso), and two are located in the new economic frontier of the northern regions (Roraima and Amapá). We found a negative correlation between human development growth and human development status at the beginning of our study period at the regional level; furthermore, this correlation was statistically significant under our null model. In fact, the correlation between human development levels during the two time periods was, of course, positive, though the SAR coefficient was significantly smaller than one (i.e., 0.787). This finding supports the hypothesis that municipalities with a low human welfare index exhibit more rapid growth than do municipalities with a high human welfare index (β-convergence), as was proposed by Caviglia-Harris et al. (2016). Also, differences in 140

Land Use Policy 65 (2017) 135–142

J.M.C.d. Silva et al.

2016). However, protected areas and indigenous lands across the Brazilian Amazonia remain largely unfunded and ill-protected (Ricardo and Ricardo, 2011; WWF-Brasil and ICMBIO, 2012; Araújo and Barreto, 2015) despite the fact that setting public areas aside to conserve biodiversity and protect traditional as well as indigenous land rights have been recognized as the most effective means of controlling regional deforestation (Silva et al., 2005; Garda et al., 2010; SoaresFilho et al., 2010). Our study demonstrates the following with regard to changes in human welfare across various municipalities in the Brazilian Amazonia from 2005 to 2012: (a) signals of a boom-and-bust pattern are weak, and therefore this pattern cannot be generalized across the entire region; (b) human welfare is increasing more rapidly in low-development municipalities than in high-development cities, and all municipalities are converging on at least one regional average rather than on a national average; (c) urbanization does not lead to positive changes in human welfare, which indicates that the infrastructure available in regional urban centers is limited; (d) public investments are negatively associated with human welfare growth, thus suggesting that if public investments are not used to leverage the potential of other sectors of the local economy, human welfare will not improve; (e) agriculture is negatively associated with positive changes in human welfare at the local level, possibly due to the dominance of cattle-ranching as the predominant economic activity of this sector; and (f) state-level policies matter, and further analyses of regional trends in the realm of development and conservation across the region should take these policies into account. Finally, we suggest that although human welfare and deforestation have a weak statistical relationship, we cannot contend that they have been fully decoupled. Forest loss across the region is still pervasive (INPE, 2015), and institutions are too weak to sustain the transition from a frontier development model to a conservation-centered plan (Aguiar et al., 2016). A total decoupling of improvements in human welfare and deforestation across the Brazilian Amazonia is a feasible target (Vieira et al., 2005), though it requires the adoption of an integrated regional development strategy that promotes the intensification and diversification of economies in areas that have already been deforested; this strategy must be applied alongside the outright conservation and sustainable use of all remaining forests and the implementation of programs that reduce existing regional social inequalities (Becker, 2004; Garda et al., 2010; Silva, 2015; Nobre et al., 2016).

as we found a significant but negative relationship between the two variables. The same pattern was found in five of the nine states. In Mato Grosso—the state whose municipalities enjoy, on average, the largest contributions from the agricultural sector to their GDPs—the relationship was negative but not significant. This negative relationship between human welfare and agriculture is possibly a consequence of the regional dominance of the cattle-ranching industry over all other economic activities that compose the agricultural sector. In fact, 62% of the lands that were deforested in the Brazilian Amazonia are occupied by pastures (Aguiar et al., 2016), some of which were opened only as a means of establishing property rights in the absence of formal land markets (Merry et al., 2008). Across the region, cattle-ranching is considered to be an economic activity that typically occupies large areas, suffers from low productivity, generates few jobs, and drives social conflicts (Arima et al., 2005; Prates and Bacha, 2011; Souza et al., 2015). In the last two decades, there has been a trend toward the modernization of the region’s agricultural sector, as demonstrated by the adoption of technologies aimed at increasing productivity while reducing the consumption of natural resources; furthermore, efforts have been made to replace cattle-ranching lands with croplands or, at least, with integrated systems that combine pastures and croplands (Galford et al., 2013). However, efforts to modernize the sector are not without their problems. Agricultural intensification can exarcebate water and soil pollution via the introduction of chemicals and fertilizers that can cause health problems in downstream populations (Gomes and Barizon, 2014). Additionally, not everyone can access these new technologies and practices. Rather, the technologies are restricted to those with financial resources, political influence, and investment incentives. Therefore, while the trend toward modernization is important because it can help to reduce the deforestation of private lands (Galford et al., 2013), it can also cause unintended and serious social problems. The concentration of efficient agricultural systems in the hands of a few individuals or corporations can lead to a weakening of traditional small farm production systems, high rural unemployment rates, social inequality, and health problems caused by chemical pollution (Prates and Bacha, 2011; Aguiar et al., 2016). These four factors together can reduce rather than increase average local human welfare. Caviglia-Harris et al. (2016) suggested that human development and deforestation across the region has been decoupled. However, full decoupling means that deforestation ceases while human welfare indicators continue to increase. Although zero deforestation is a relatively longstanding goal for the Brazilian Amazonia (e.g., Vieira et al., 2005), forest loss continues to be pervasive across the region (INPE, 2015). There has been some recent excitement about the potential of public-private interventions in individual supply chains of commodities such as soybeans and beef to achieve zero deforestation across the region, though the results thus far have been limited in comparison to the magnitude of the actual problem (Gibbs et al., 2016). Macedo et al. (2012) found that among municipalities in Mato Grosso, soy production—a major agricultural commodity—has become decoupled from deforestation as a consequence of the adoption and enforcement of modern policies and technologies. However, we believe that it would be premature to claim that full decoupling between deforestation and human welfare has been achieved in Mato Grosso, as the state is still one of the regional leaders in forest loss and social conflicts (INPE, 2015; Souza et al., 2015). Additionally, individual municipalities within a state can reverse the decoupling process if they do not diversify their economies over time, if the prices of agricultural commodities decrease, or if forest conservation policies are relaxed. The full decoupling of deforestation and human welfare could be most easily achieved in those municipalities wherein most of their territory is located within protected areas or indigenous lands (e.g., several municipalities in Amapá, Pará, Amazonas, and Roraima), assuming that these areas are managed well enough to fulfill their goals as well as generate sufficient revenue to foster strong local economies (Dias et al.,

Acknowledgements JMCS was partially funded by a generous gift from the Swift Action Fund. JMCS and SP received support from the University of Miami, and JAFDF was funded by the Brazilian National Council for Scientific and Technological Development (CNPq). We thank Fábio Scarano and an anonymous reviewer for providing useful suggestions on earlier draft of the manuscript. References Aguiar, A.P.D., Vieira, I.C.G., Assis, T.O., Dalla-Nora, E., Toledo, P.M., Santos-Junior, R.A.O., Batistella, M., Coelho, A.S., Savaget, E.K., Aragão, L.E.O.C., Nobre, C.A., Ometto, J.P.H., 2016. Land use change emission scenarios: anticipating a forest transition process in the Brazilian Amazon. Glob. Change Biol. 22, 1821–1840. Araújo, E., Barreto, P., 2015. Estratégias e Fontes de Recursos para Proteger as Unidades de Conservação da Amazônia. Imazon, Belém. Arima, E., Barreto, P., Brito, M., 2005. Pecuária na Amazônia: Tendências e Implicações para a Conservação Ambiental. Imazon, Belém. Assunção, J., Gandour, C., Rocha, R., 2015. Deforestation slowdown in the Brazilian Amazon: prices or policies? Environ. Dev. Econ. 20, 697–722. Barberia, L.G., Biderman, C., 2010. Local economic development: theory, evidence, and implications for policy in Brazil. Geoforum 41, 951–962. Barro, R., Sala-i-Martin, X., 1995. Economic Growth. McGraw-Hill, New York. Becker, B.K., 2001. Revisão das políticas de ocupação da Amazônia: é possível identificar modelos para projetar cenários? Parcerias Estratégicas 12, 135–159. Becker, B.K., 2004. Amazônia: Geopolítica na Virada do III Milênio. Editora Garamond,

141

Land Use Policy 65 (2017) 135–142

J.M.C.d. Silva et al.

147–151. Legendre, P., Legendre, L., 1998. Numerical Ecology. 2nd English Edition. Elsevier, Amsterdam. Macedo, M.N., DeFries, R.S., Morton, D.C., Stickler, C.M., Galford, G.L., Shimabukuro, Y.E., 2012. Decoupling of deforestation and soy production in the southern Amazon during the late 2000. Proc. Natl. Acad. Sci. U. S. A. 109, 1341–1346. Mather, A., 1992. The forest transition. Area 24, 367–379. Merry, F., Amacher, G., Lima, E., 2008. Land values in frontier settlements of the Brazilian Amazon. World Dev. 36, 2390–2401. Nobre, C.A., Sampaio, G., Borma, L.S., Castilla-Rubio, J.C., Silva, J.S., Cardoso, M., 2016. Land-use and climate change risks in the Amazon and the need of a novel sustainable development paradigm. Proc. Natl. Acad. Sci. U. S. A. 113, 10759–10768. Perz, S.G., 2000. The quality of urban environments in the Brazilian Amazon. Soc. Indic. Res. 32, 25–60. Prates, R.C., Bacha, C.J.C., 2011. Os processos de desenvolvimento e desmatamento da Amazônia. Economia e Sociedade 20, 601–636. R Development Core Team, 2015. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna. Rajão, R., Georgiadou, Y., 2014. Blame games in the Amazon: environmental crises and the emergence of a transparency regime in Brazil. Glob. Environ. Polit. 14, 97–115. Ricardo, B., Ricardo, F., 2011. Povos Indígenas no Brasil 2006/2010. Instituto Socioambiental, São Paulo. Richards, P., VanWey, L., 2015. Where deforestation leads to urbanization: how resource extraction is leading to urban growth in the Brazilian Amazon. Ann. Assoc. Am. Geogr. 105, 806–823. Rodrigues, A., Ewerrs, R., Parry, L., Souza Jr., C., Veríssimo, A., Balmford, A., 2009. Boom-and-bust development patterns across the Amazon deforestation frontier. Science 324, 1435–1437. Rudel, T.K., Coomes, O.T., Moran, E., Achard, F., Angelsen, A., Xu, J., Lambin, E., 2005. Forest transitions: towards a global understanding of land use change. Glob. Environ. Change 15, 23–31. Santos, D., Celentano, D., Garcia, J., Aranibar, A., Veríssimo, A., 2014. Índice de Progresso Social na Amazônia Brasileira: IPS Amazônia 2014. Imazon e Social Progress Imperative, Belém. Schneider, R., Arima, E., Veríssimo, A., Barreto, P., Souza Jr., C., 2002. Sustainable amazon: limitations and opportunities for rural development. World Bank Technical Paper No. 515, Environmental Series. World Bank, Washington DC. Silva, J.M.C., Rylands, A.B., Fonseca, G.A.B., 2005. The fate of the Amazonian areas of endemism. Conserv. Biol. 19, 689–694. Silva, J.M.C., 2015. A conservação da biodiversidade como estratégia competitiva para a Amazônia no antropoceno. In: Silva, O.M.A., Homma, A.K.O. (Eds.), Pan-Amazônia: Visão Histórica Perspectivas de Integração e Crescimento. FIEAM, Manaus, pp. 140–156. Soares-Filho, B., Moutinho, P., Nepstad, D., Anderson, A., Rodrigues, H., Garcia, R., Dietzsch, L., Merry, F., Bowman, M., Hissa, L., Silvestrini, R., Maretti, C., 2010. Role of Brazilian Amazon protected areas in climate change mitigation. Proc. Natl. Acad. Sci. U. S. A. 107, 10821–10826. Souza, P.F., Xavier, D.R., Rican, S., Matos, V.P., Barcellos, C., 2015. The expansion of the economic frontier and the diffusion of violence in the Amazon. Int. J. Environ. Res. Public Health 12, 5862–5885. Vieira, I.C., Silva, J.M.C., Toledo, P.M., 2005. Estratégias para evitar a perda de biodiversidade na Amazônia. Estudos Avançados 19, 153–164. WWF-Brasil, ICMBIO, 2012. Efetividade da Gestão das Unidades de Conservação Federais do Brasil: Resultados de 2010. WWF-Brasil and ICMBIO, Brasília. Weinhold, D., Reis, E.J., Vale, P.M., 2015. Boom-bust patterns in the Brazilian Amazon. Glob. Environ. Change 35, 391–399.

Rio de Janeiro. Bivand, R., Pebesma, E., Gómez-Rubio, V., 2013. Applied Spatial Data Analysis with R, 2nd edition. Springer, New York. Burnham, K.P., Anderson, D.R., 2002. Model Selection and Multimodel Inference: A Practical Information–Theoretical Approach. Springer, New York. Castro, J.V., 2004. Indicators of Real Economic Convergence: A Primer. UNU-CRIS eWorking Papers. http://cris.unu.edu/indicators-real-economic-convergence-primer. (Accessed 24 May 2016). Caviglia-Harris, J., Sills, E., Bell, A., Harris, D., Mullan, K., Roberts, D., 2016. Busting the boom-bust pattern of development in the Brazilian Amazon. World Dev. 79, 82–96. Celentano, D., Sills, E., Sales, M., Veríssimo, A., 2012. Welfare outcomes and the advance of the deforestation frontier in the Brazilian Amazon. World Dev. 40, 850–864. Clement, C., Higuchi, N., 2006. A floresta amazônica e o future do Brasil. Ciência e Cultura 58, 44–49. Cressie, N.A.C., 1993. Statistics for Spatial Data Analysis. John Wiley and Sons, New York. Dias, T.C.A.C., Cunha, A.C., Silva, J.M.C., 2016. Return on investment of the ecological infrastructure in a new forest frontier in Brazilian Amazonia. Biol. Conserv. 194, 184–193. FIRJAN, 2015. Índice FIRJAN de Desenvolvimento Municipal (IDFM). www.firjan.com. br/ifdm. (Accessed 24 May 2016). Fearnside, P.M., 1997. Environmental services as a strategy for sustainable development in rural Amazonia. Ecol. Econ. 20, 53–70. Fortin, M.J., Dale, M., 2005. Spatial Analysis: A Guide for Ecologists. Cambridge Univ. Press, Cambridge. Galford, G.L., Soares-Filho, B., Cerri, C.E.P., 2013. Prospects for land-use sustainability on the agricultural frontier of the Brazilian Amazon. Phil. Trans. R. Soc. B 368, 20120171. Garda, A.A., Silva, J.M.C., Baião, P., 2010. Biodiversity conservation and sustainable development in the Amazon. Syst. Biodivers. 8, 169–175. Gibbs, H.K., Ruesch, A.S., Munger, J., Chelly, I., Morton, D.C., Noojipady, P., SoaresFilho, B., Barreto, P., Micol, L., Walker, N.F., 2015. Brazil’s soy moratorium. Science 347, 377–378. Gibbs, H.K., Munger, J., L’Roe, J., Barreto, P., Pereira, R., Christie, M., Amaral, T., Walker, N.F., 2016. Did ranchers and slaughterhouses respond to zero-deforestation agreements in the Brazilian Amazon. Conserv. Lett. 9, 32–42. Gomes, M.A.F., Barizon, R.R.M., 2014. Panorama da Contaminação Ambiental por Agrotóxicos e Nitrato de Origem Agrícola no Brasil: Cenário 1992/2011. Embrapa Meio Ambiente, Jaguariúna. Haining, R., 2002. Spatial Data Analysis. Cambridge Univ. Press, Cambridge. Hansen, M.C., Potapov, P.V., Moore, R., Hancher, M., Turubanova, S.A., Tyukavina, A., Thau, D., Stehman, S.V., Goetz, S.J., Loveland, T.R., Kommareddy, A., Egorov, A., Chini, L., Justice, C.O., Townshend, J.R.G., 2013. High-resolution global maps of 21st-century forest cover change. Science 342, 850–853. Hecht, S.B., 2011. From eco-catastrophe to zero deforestation? Interdisciplinarities, politics, environmentalisms and reduced clearing in Amazonia. Environ. Conserv. 39, 4–19. IBGE, 2004. Mapa de Biomas do Brasil, Escala 1:5,000,000. Instituto Brasileiro de Geografia e Estatística (IBGE), Rio de Janeiro. IBGE, 2008. Produto Interno Bruto dos Municípiosm. Série Relatórios Metodológicos, 2a edição. IBGE, Rio de Janeiro. IBGE, 2011. Sinopse do Censo Demográfico 2010. IBGE, Rio de Janeiro. IBGE, 2015. Produto Interno Bruto dos Municípios 1999–2013. IBGE. http://www.sidra. ibge.gov.br/bda/pesquisas/PIBMun. (Accessed 24 May 2016). IBGE, 2014. Produto Interno Bruto dos Municípios 2012. IBGE, Rio de Janeiro. INPE, 2015. Projeto PRODES: Monitoramento da Floresta Amazônica Brasileira por Satélite. INPE. www.obt.inpe.br/prodes. (Accessed 24 May 2016). Jackson, D.A., Somers, K.M., 1991. The spectre of ‘spurious’ correlations. Oecologia 86,

142