Which Amazon problem? Problem-constructions and transnationalism in Brazilian presidential discourse since 1985

ABSTRACT

The Amazon is a complex object of policy that comprises environmental, economic, social, and sovereignty concerns. Despite this complexity, governments are often portrayed as having a single understanding of the region as a political problem. In this article, we investigate how the Amazon has been constructed as a problem in 6240 Brazilian presidential speeches since 1985 using supervised machine learning. Conceptually, we develop a framework that accounts for how important transnational actors, such as presidents, construct objects of policy as particular problems depending on where and when they participate in politics. Empirically, we find that presidents often construct the Amazon as an environmental problem when speaking far away from the region, whereas they usually construct it as problems of economic integration or social development when speaking within in the Amazon. Furthermore, presidents increasingly mix problem-constructions to represent the Amazon as a complex and multifaceted object of policy.

KEYWORDS:

• Discourse analysis
• presidential speeches
• transnational governance
• environmental policy
• Brazilian Amazon
• supervised machine learning

1. Introduction

The Amazon Rainforest extends across eight international borders. The Brazilian Amazon covers around 60% of the whole biome and is home to over 28 million people. Its most inhabited city, Manaus, enjoys a distinct fiscal regime designed to create jobs in the region and connect its economy to the rest of the country. An extensive portion of the Brazilian Amazon is protected by a combination of indigenous territories and protected areas, rendering the Rainforest an important component of global climate mitigation. The Brazilian Amazon is, thus, a complex object of policy comprising environmental, economic, social, and sovereignty concerns. Despite this complexity, the literature about policies in the Amazon often portrays governments as having a single understanding of what the Amazon problem is. Although this literature is helpful to compare variation across different governments, such analyses oversimplify policies within governments and misrepresents what, in reality, is more intricate and inconsistent. In this article, we investigate how the Amazon has been constructed as a problem in Brazilian presidential speeches since 1985.

We combine Bacchi (1999)’s theory of problem representation and Keck and Sikkink (1998)’s theory of transnationalism to develop a framework outlining how transnational actors, as presidents, construct objects of policies as particular problems depending on where and when they participate in politics. To examine this variation, we create a dataset containing 6240 official speeches by all Brazilian presidents since 1985. Drawing from the socio-cultural history of the Amazon, we inductively develop a codebook to categorize problems related to political concerns over the Amazon. These problem-constructions touch upon national sovereignty, economic integration, social development, or environmental conservation. We employ supervised machine learning to classify how Amazon related statements construct the Amazon as a problem. We then conduct an inferential and descriptive analysis of the data. A systematic focus on presidential speeches is important because scholars recognize decisions to deforest on the ground are derived from government’s policies and discourse (Campbell 2015, Capobianco 2021). Still, we lack empirical accounts of the Amazon in presidential discourses.

We have four main findings. First, when presidents are far away from the Amazon region, they are more likely to construct the Amazon as an issue of environmental conservation. Conversely, within the Amazonian region, presidents usually construct it as an issue of social development and economic integration. Second, presidents are more likely to construct the Amazon as an environmental conservation problem when deforestation rates decrease. When deforestation rates increase, presidents tend to side-step the issue and construct the Amazon as a problem of economic integration. Third, we find that high-profile environmental events which receive strong international media attention, drive up the frequency at which the Amazon appears in discourses. However, these events do not systematically change how the Amazon is constructed as a problem. Finally, we find that presidents are increasingly more likely to mix problem-constructions and construct the Amazon as a complex and multifaceted object of policy.

Empirically, we contribute to the literature about the Amazon by providing the first overview of where, how, and when the Amazon has been constructed as a problem in Brazilian presidential speeches. Other analyses of Brazilian environmental discourse and policies do not adopt methodological strategies distinguishing how trasnational actors behave differently inside and outside the country. We show how the same government can construct the Amazon as a different problem depending on where and when presidents speak. Conceptually, a focus on problem-constructions highlights the extent to which Amazonian problems are ignored or privileged in transnational politics and avoids assuming that the Amazon is solely an environmental problem.

We divide this article into four sections. First, we outline the connection between problem-constructions, transnationalism, and presidential speeches. Next, we present our research design including the description of our data and methodological approaches. In the third section, dedicated to the analysis, we display the results of our inferential models and move to explain developments in problem-constructions over time. We conclude by summarizing the empirical and conceptual contributions of the article.

2. Theory: problem-construction and transnationalism

2.1. Problem-representation in presidential discourse

Scholars writing about policies in the Brazilian Amazon often describe federal governments as proponents of a cohesive set of policies toward the Amazon (see Drummond and Flavia Barros-Platiau 2006, Pádua 2012, Franchini and Viola 2019, Capobianco 2019, Pereira and Viola 2021). The 1964 military dictatorship, for example, is associated with securing sovereignty in the Amazonian region by populating and integrating it into the national economy (Drummond and Flavia Barros-Platiau 2006). Whereas governments from the late-1980s up to the late-2000s are associated with a turn towards policies that focus on environmental conservation of the Amazon Rainforest (Pádua 2012). The presidencies of Rousseff (2011–2016) and Temer (2016–2018) are connected to the de-prioritizing of environmental conservation, while Bolsonaro (2019–2022) with the dismantling of environmental policies (Capobianco 2019). While these works are important to understand how different governments act towards the Amazon, they represent governments monolithically by bundling them into cohesive policy cycles.

Elsewhere, scholars have shown that there is more variation and contradiction within a specific government’s policies than suggested by the literature above (see Fairfield and Garay 2017, Ponce de Leon 2021, López 2023). We argue that the reason behind these monolithic representations of governments is conceptual: many studies about the region contain an explicit or implicit assumption of what the ‘Amazon problem’ is. Because of the Amazon Rainforest’s role in global climate-mitigation, it is often assumed to be an environmental problem that demands environmental solutions. The Amazon region, its peoples, and forests, though, are a historically complex object of policies that also comprise economic, social, and sovereignty concerns.

Bacchi (1999, p. 63) places the question of ‘what is the problem represented to be’ in the center of policy studies. The existence or proposal of a policy implies that there is a (public) problem that needs (governmental) action to be fixed. Hence, discourse represents, implicitly or explicitly, objects of policies as instances of a specific problem. Problem-representation matters because it carries implications related to who is involved both in the root causes and in the solutions of the problem (Bacchi 1999). It also helps us shed light on which problems are ignored or privileged. Thus, we place discourse about policy objects at the center of our conceptual framework. We define policy objects as objects that demand dedicated political attention and argue that by speaking about a policy object, governments are making a deliberate choice to represent, or construct it, as a specific problem. The problem might be either explicitly stated or implied in the proposed solution. This is what we label as problem-constructions.

Presidential discourses are a good instance to identify problem-constructions over time and across spaces. Presidential speeches and policy share a co-constitutive relationship as speeches have the power to introduce, define, and justify public policy, as well as shape issue perception to broad audiences (Zarefsky 2004). This is particularly true in the Brazilian context where the executive branch of government, led by the president, has historically been the primary actor in shaping public policy, especially after the adoption of the 1988 constitution (Macaulay 2017). In Brazil’s hybrid system, often referred to as a Coalitional Presidentialism (Couto et al. 2021), presidents build legislative support for a broad agenda through the distribution of positions in the state bureaucracies. They also have vast decree powers to adopt policy unilaterally (Carlos et al. 2008). This means that presidents often have enough power and support to implement the agenda they deem appropriate, especially in the case of climate change (Hochstetler 2021).

2.2. Settings, urgency and variation in problem-constructions

There are two main theories for explaining variation in problem-construction in international/transnational politics are: Putnam’s two-level game and Keck and Sikkink’s transnational networks.

Putnam (1988) argues that the outcomes of international negotiations lie within the overlap between the interests of domestic and international groups. At the national level, interest groups pressure the government to adopt favorable policies. Internationally, the government seeks to fulfill domestic preferences ‘while minimizing the adverse consequences of foreign developments’ (Putnam 1988, p. 434). The theory has been expanded to cover multi-level games (see Collinson 1999), but its core remains the same: the outcome of negotiations should lie within the overlap of distinct interest groups. Consequently, presidential speeches are not necessarily contradictory across venues but a common denominator of the interests of different groups. Adopting two-level games to explain American presidential speeches about climate change, Calderwood (2020) found little support that location changes the content of presidential speeches.

The two-level game theory implies that actors respond rationally to systemic pressures imposed by their position in relation to interest groups. This view seems incompatible with the possibility of varied problem-constructions for the same policy object. Rather, we start from the idea of transnationalism being relationships that transcend nation-states and incorporate non-state actors, from the local level to the international level (Nye and Keohane 1971, Keck and Sikkink 1998). In this context, international summits, for example, are considered transnational spaces because they bring together international media, corporations, non-governmental organizations, social movements, and various state actors.

Keck and Sikkink (1998)’s theory of transnational networks rejects the dichotomy of international and domestic levels and argues that the identity and goals of transnational actors are not derived from their structural location vis-a-vis domestic and foreign interest groups. Rather, international politics as transnational networks have ‘a structured and structuring dimension’ (Keck and Sikkink 1998, p. 4): state and non-state actors participate in and shape international politics. Although their theory is developed in the context of value-based advocacy groups (i.e. non-state actors), the overall view of international politics holds for state and other non-state actors.

Important participants of transnational networks, as presidents, can use a myriad of tactics to introduce, define, persuade, socialize, and pressure when interacting with other actors. They can create politically meaningful information and move it where they believe it can have an impact or leverage institutions to realize symbolic or material gains (Keck and Sikkink 1998). In this sense, the positions of actors in transnational networks derive from complex interactions in which their identities and goals are malleable and can vary depending on the space and time of the interaction.

Problem-constructions are choices transnational actors have of how to represent objects of policies as problems. If malleable identities across space and time are a defining feature of transnational politics, it follows that transnational actors’ choice of problem-constructions depend on their views of an audience at a certain period. Specifically, how pressing a problem is in a certain space or at a certain time.

Space

Transnational actors construct problems and attempt to inform or leverage audiences for symbolic and material gains according to their expectations of that audience. We conceptualize space in terms of settings. Settings are levels of aggregation for transnational audiences that make sense in light of the object of policy under study. This indicates how transnational actors construct policy objects as problems can vary for instance, within the country – speaking inside the Amazon region versus outside – because transnational actors identify these audiences as distinct. The same holds true when speaking abroad – in Amazonian countries, non-Amazonian countries, and at international summits – where transnational actors’ expectations and the identities of an audience could diverge completely from that of domestic actors. In each one of these settings transnational actors’ expectations might influence how a problem is implicitly or explicitly constructed.

Time

How transnational actors construct policy objects varies as the urgency of certain problems changes and new information arises. We conceptualize time as the urgency of broad and related issues to those surrounding a policy object. This is largely defined by exogenous factors to the speaker. This indicates, for instance, that as climate change becomes more urgent and mainstream as a broad issue, it is more likely that actors will consider mentioning climate change, when discussing various forms of public policy. For instance, when speaking about subsidies to fossil fuel companies in the 1980s, presidents would likely not mention the dangers of climate change. In a similar scenario today, climate change would likely be mentioned given the increased relevance of the topic.

While time and space interact, here we are interested in their cumulative effects. For example, when speaking in the 1990s to an audience that finds climate change pressing, transnational actors would likely mention it as a problem independently of the perceived urgency of a climate breakdown at the time. Accordingly, transnational actors might even consider touching upon the issue while speaking at a business association today, because of its urgency in time. If our conceptual framework holds, it entails that transnational actors view different settings as distinct policymaking instances. Although there might exist convergences of priorities over time because problems become more urgent, contradictions are possible as settings change. This does not mean that policies and priorities change instantaneously and consequently do not matter. Rather, transnational actors will focus on representing issues based on what they perceive as possible symbolic or material gains from the audience.

Our conceptual framework, thus, combines Bacchi’s theory of problem-representation and Keck and Sikkink’s theory of transnationalism to account for diverse problem-constructions in presidential discourses. We help to advance the literature about the Amazon in three ways. First, we provide an outlook of the region, its people, and forest beyond an environmental concern. In doing so, we take a step back and avoid assumptions what type of problem the Amazon is. Instead, we focus on how the Amazon has been represented as a problem. Second, other analyses of Brazilian environmental policies center on its national and international dimensions without adopting methodological strategies that distinguish how important actors behave differently in both policymaking instances. Our focus on presidential speeches affords us the analytical leverage to identify transnational variation, as speeches take place in multiple settings over time. Finally, presidential speeches in Brazil have been studied for topics such as race relations (Da Silva et al. 2019) and inequality (Grangeia 2017). When it comes to the environment, a few recent studies focus primarily on Bolsonaro’s discourse (Marquardt et al. 2022, Mendes Motta and Hauber 2022). We argue, however, that a longitudinal view across Brazilian presidents is important to understand how much continuity and change there is.

3. Research design

3.1. Data

We update Cezar’s (2020) dataset of all Brazilian presidential speeches from 1985 to 2019 to include speeches from 2020 to 2022. The final dataset contains 6240 speeches with their date and state or country location. We proceed to identify speeches that refer to the Amazon as a region, people, or forest, by detecting the stem ‘amazon’. In Portuguese, the stem captures terms such as ‘Amazonia’, ‘Amazonica’, ‘Amazonidas’, ‘Amazonense(s)’, ‘Amazonas’, among others. We find that 968 speeches are, at least partially, about the Amazon. We use the poldis R package (Sposito 2021) to extract two sentences before and after each reference to the Amazon in the presidential speeches. We opt for picking two sentences around, rather than a number of words, because sentences usually contain a cohesive idea. These sentences form our unit of analysis: 2048 unique Amazonian statements, with an average length of 131 words.

3.2. Operationalizing Amazonian problem-constructions

We use a codebook, developed through an inductive process and informed by the literature, to categorize Amazonian statements into four problem-constructions: national sovereignty, economic integration, environmental conservation, and social development. While the three former ones are present in the literature, we argue a fourth one exists: social development. Amazonian statements can construct the Amazon as a single problem (i.e. pure-type statements), or can construct the amazon as referring to multiple problem-constructions (i.e. mixed-types), as Figure 1 (below) details.

Figure 1. Operationalization of problem-constructions.

3.2.1. National Sovereignty

In the process of securing Amazonian borders since the 18th century, Brazil thwarted ‘the imperial ambitions of France, Britain, the United States, Belgium, Bolivia, and Peru’ (Hecht and Cockburn 1990, p. 8), and secured most of the Amazon under its territory. While military diplomacy was very successful, the process did not come without its issues, and sovereignty in the Amazon became one of the main military concerns (Rajão et al. 2020). More recently, as we move to a world where non-state actors gain importance in international politics (Keck and Sikkink 1998, Andonova 2014), sovereignty-related problems become varied. Threats to national sovereignty, consequently, broaden from nation-states to a wide set of actors. The sovereignty problem-construction advances the view that the Brazilian Amazon belongs to Brazil and foreign or non-state presence in the region are concerning. The policy solutions that implicitly represent this problem include the monitoring of the borders and strict regimes related to entry into the region.

3.2.2. Economic integration

The Vargas dictatorship (1937–46) and the military dictatorship (1964–89) took over the task of modernizing the Amazon (Becker 2005). In 1966, the Brazilian Military launched ‘Operation Amazon’, a policy to modernize the region based on three assumptions (Acker 2021). First, it assumed that nature should be conquered by mankind. Second, it assumed that exploiting natural resources would render the Amazon region economically profitable. Third, it assumed that populating the region was necessary to integrate it into the country and exert control over the territory. Concretely, this meant a series of major infrastructure projects, incentives for settlers to expand the agricultural frontier, and the establishment of a tax-free zone in Manaus to attract investments. The economic integration problem-construction advances the view that the Brazilian Amazon needs to be developed and modernized. These policy solutions have at their core the development of the necessary infrastructure (physical and fiscal) to integrate the region into the national and international economy.

3.2.3. Environmental conservation

The rapid economic changes in the region from the 1960s to the 1980s were matched with the birth of domestic environmental institutions (Drummond and Flavia Barros-Platiau 2006). The creation of these institutions stemmed from the impression of the lack of control over the environment, following the efforts to economically integrate in the region (Hecht and Cockburn 1990, Acker 2021, Capobianco 2021). This process accelerated in the late 1980s with new forms of environmentalism (Viola 1987). The environmental conservation problem-construction emphasizes that the Amazon should be preserved by the creation of protected areas and deforestation should be halted. The policy solutions imply further investments in command-and-control institutions, the valuation of standing ecosystems through incentive schemes, and the creation of protected areas.

3.2.4. Social development

As democracy established itself in Brazil, especially with the 1988 constitution, settlers in the Amazon became electorates with constitutional rights. Governments can emphasize the lack of hospitals, sanitation, and schools concerning peoples’ dignity, standards of living, and other constitutional rights. However, such emphasis can be masked within both the economic integration and environmental conservation accounts. While for some groups, such as indigenous peoples, constitutional rights and preservation of the forest go hand in hand, providing better standards of living to large portions of populations of the Amazonian states and conservation are not necessarily connected. Social development is a problem-construction that focuses on constitutional rights, citizenship, and dignity. The policy solutions imply enabling access to water, sanitation, electricity, internet, and radio, as well as the construction of schools and hospitals.

With the codebook in hand, we then randomly selected a sample of 1024 Amazonian statements (50% of our data) and each author hand-coded these separately. Once manual coding was finished, the authors resolved disagreements through discussion and refined the codebook (in supplementary material). We then randomly divided the hand-coded data into a training set (80% of observations) and validation set (20% of observations) to train a machine learning algorithm. We chose to hand code half of the observations because there are multiple categories of how presidents talk about the Amazon, and the training set size should be increased (see Justin et al. 2022).

We use a support-vector machine algorithm, a non-probabilistic linear classifier appropriate for binary categories (Noble 2006, Meyer et al. 2021), to train a model to label text. The model was validated by testing it on the validation set. As a robustness check, the authors randomly select 100 automatically coded statements and find that the automated model is 95% accurate in coding these statements as the authors would. The final dataset, excluding false positive matches, contains 1924 coded Amazonian statements.¹ Automating the code saved the authors over one month of work.

3.3. Operationalizing settings and urgency

Our explanation for variation in problem-constructions is based on the setting of a speech and the urgency of issues. This comprises our independent variables, as presented below.

3.3.1. Settings

To operationalize settings, we extract the location of the speech and create four distinct settings: Amazonian states within Brazil, non-Amazonian states within Brazil, Amazonian countries, and non-Amazonian countries. Each one of these settings encompasses audiences that have different expectations in reference to our policy object, the Amazon. When speaking in Amazonian states presidents face audiences that likely have distinct social, economic, or environmental demands in relation to the Amazon. Such expectations might be different from the ones’ audiences in non-Amazonian states within Brazil have. Furthermore, during visits to other Amazonian countries, presidents may perceive them as peers facing similar challenges related to the Amazon and adjust their speeches accordingly. Conversely, when speaking in non-Amazonian countries (e.g. international summits or bilateral visits), presidents may tailor their speeches to align with what they perceive as possible symbolic or material gains from the audience.

This operationalization of settings comes with limitations. While we would ideally like to capture the exact venue of the speech (i.e. multilateral mechanism, private fora, bilateral visit, the inauguration of a bridge or road), it is difficult to programmatically code exact locations from thousands of speeches that often mention multiple locations in the same text. Nevertheless, we believe the four settings coded are enough to demonstrate whether, and how, problem-construction changes transnationally.

3.3.2. Urgency

To capture the urgency of issues, we use two proxies: lagged deforestation rates in the Brazilian Amazon and annual inflation rates. When deforestation rates are high, the problem of environmental conservation can be perceived as more urgent. Relatedly, when inflation is high, the problem of economic integration issues can be perceived as more urgent. We lag deforestation rates by one year because official deforestation rates are usually published in the following year. Since both annual deforestation rates and annual inflation rates come in different units of measurement (i.e. percentage and kilometers square) we scale these values to facilitate interpretation of coefficients. Scaling variables does not affect statistical inference in regression models.

This approach to operationalizing urgency also has limitations. Ideally, we would like to add proxies of urgency for social and sovereignty issues. However, social development indicators (e.g. life expectancy, education levels, access to sanitation) generally increased over time in Brazil. In terms of national sovereignty, the way we conceptualize the problem-construction entails internal and external threats which are not easily measured. Therefore, we limit our analysis of urgency to environmental and economic issues.

3.4. Inferential and descriptive strategies

We use fixed-effect logistic regression models to predict the probability of a speech mentioning the Amazon and each of the four problem-constructions. The fixed-effects models are indexed by the president, since each president might have individual-specific characteristics that can influence outcomes. Both presidents’ political ideology and setting preferences could affect problem-construction.² Fixed-effects models help to control for these possible individual-specific issues and other unit-unvarying characteristics (Allison 2009).

The dependent variables of the models are a binary variable that takes the value of 1 if a speech mentions the Amazon (model 1); and binary variables that take the value of 1 if the Amazonian statement constructs the Amazon as a specific problem-construction (models 2–5). We chose to model each construction separately, and to focus on pure-types (i.e. statements that construct the Amazon as a single problem), to provide unambiguous interpretations and more robust findings. Our main independent variables are settings and urgency, as described above. We add one control, election years, due to the possibility of presidents constructing the Amazon in systematically different ways during campaigns. It is important to outline that we do not make any causal claims. Many scholars inquire and find causal relationships between discourse and outcomes. Our research design and data do not allow us to do so.³

4. How has the Amazon been constructed as a problem?

4.1. The Amazon problem: setting and urgency

Following our framework, we argue that presidents’ expectations about an audience in certain settings, as well as the urgency of specific issues, explain variations of when the Amazon is a problem and what type of problem it is. We start by first scrutinizing the relationship between the setting and problem-constructions. The first model (1) addresses the question of when the Amazon is a problem to begin with (Table 1). We model the probability of a speech generally mentioning the Amazon in relation to the reference category, Amazonian states within Brazil. The findings are intuitive: the likelihood of mentioning the Amazon in speeches decreases as presidents move away from the region.

Table 1. Log odds for fixed-effects logistic regressions indexed by president.

	Amazon Speech	Environmental Conservation	Economic Integration	Social Development	National Sovereignty
	(1)	(2)	(3)	(4)	(5)
Amazonian Countries	−0.167^***(0.033)	0.012(0.032)	0.091^**(0.039)	−0.097^***(0.028)	−0.004(0.022)
Non- Amazonian States	−0.403^***(0.025)	0.079^***(0.018)	−0.039^*(0.022)	−0.013(0.016)	0.028^**(0.012)
Non-Amazonian Countries	−0.430^***(0.026)	0.202^***(0.031)	−0.056(0.037)	−0.057^**(0.027)	−0.004(0.021)
Deforestation	−0.075^***	−0.098^**	0.303^***	−0.045	−0.030
	(0.025)	(0.047)	(0.056)	(0.041)	(0.031)
Inflation	0.232^***	0.164^**	−0.291^***	0.072	0.102^**
	(0.049)	(0.071)	(0.086)	(0.063)	(0.048)
Election Year	−0.010	0.041^*	0.011	0.008	−0.016
	(0.012)	(0.022)	(0.026)	(0.019)	(0.014)
Observations	6,167	1,924	1,924	1,924	1,924
R2	0.066	0.035	0.031	0.010	0.007
Adjusted R^2	0.064	0.028	0.025	0.004	0.0001
F Statistic	72.132^*** (df = 6; 6153)	11.499^*** (df = 6; 1910)	10.249^*** (df = 6; 1910)	3.320^*** (df = 6; 1910)	2.197^*** (df = 6; 1910)

Note: *p<0.1; **p<0.05; ***p<0.01.

The following models (2–5) address the question of how the Amazon is constructed as a problem in different settings (Table 1). In terms of environmental conservation (model 2), we observe that when speaking in non-Amazonian countries (e.g. in international summits or bilateral meetings) presidents are much more likely to construct the Amazon as an issue of environmental conservation, in relation to when they speak within Amazonian states in Brazil (reference category). To a lesser extent, presidents are also more likely to construct the Amazon as an issue of environmental conservation when speaking in non-Amazonian states within Brazil. While the results for economic integration (model 3) and social development (model 4) problem-constructions are not as robust, they are strong enough to suggest that presidents are less likely to construct the Amazon as problems of economic integration or social development when speaking outside of the Amazon. Lastly, presidents are more likely to construct the Amazon as an issue of national sovereignty when speaking in non-Amazonian states within Brazil.

We interpret these findings to be consistent with our conceptual framework: presidents construct the Amazon as different problems based on their expectations of settings. When it comes to non-Amazonian countries, it is likely that Brazilian presidents construct the Amazon as an issue of environmental conservation to leverage financial support for policies in the Amazonian region. The Brazilian government successfully funded domestic public policy with transnational financial support on several occasions since the 1990s, most notably with the Amazon Fund which provided over USD 1 billion for the region (Silva-Muller and Faul 2022). That is, presidents realize material and symbolic gains outside of Brazil and of Amazonian countries, by constructing the Amazon as an environmental problem.

In turn, when speaking within the Brazilian Amazon, presidents likely see their audiences as local constituencies. Presidents pursue a national agenda of economic integration and social development within Brazil while, concurrently, financing environmental policies with transnational funds. For example, rural credit offered to local agricultural producers in Amazonian states increased eight-fold from 1999 to 2012, reaching about USD 2 billion yearly (Capobianco 2021). As well, the special fiscal regime in Manaus, designed to increase industry in the region, was largely maintained. Moreover, additional roads and dams (e.g. the BR-163 highway and the Belo Monte dam and hydroelectric power plant) were constructed during the same period. Presidents highlight economic and social problem-constructions within the Amazon as they speak to voters to whom they are constitutionally accountable and who are perceived as more interested in socio-economic problems than environmental ones. We interpret this as evidence that presidents view settings as distinct policymaking instances.

We also argue that the urgency of issues in time influences the likelihood of the president mentioning the Amazon and what they will say. We find evidence across our models that this holds in unexpected ways. Presidents are less likely to generally mention the Amazon in speeches when the urgency of environmental related issues increase (i.e. rising deforestation rates). The opposite is true for inflation. As inflation increases, and so the urgency of economic problems, presidents are more likely to generally mention the Amazon in speeches (model 1). This relationship is more intricate when we look at how Amazonian constructions vary in presidential speeches as environmental or economic issues become more urgent. We find evidence that when the urgency of environmental related issues increases, the probability of presidents constructing the Amazon as an environmental conservation problem decreases (model 2) and the probability of constructing it as an issue of economic integration increases (model 3). Alternatively, as the urgency of related economic issues increases, the probability of constructing the Amazon as an issue of environmental conservation increases (model 2) and the probability of presidents constructing it as an issue of economic integration decreases (model 3). We interpret that by opting for an alternative problem-construction to an issue which is urgent, presidents side-step problems. In turn, by opting for an aligned problem-construction to an issue which is not urgent, presidents boast about positive outcomes.

4.2. Amazonian speeches over time

Figure 2, below, shows the proportion of speeches that mention the Amazon in relation to all speeches in a year. By investigating its development in time, we can identify instances when the Amazon as a general object of policy becomes a more pressing problem. On average, presidents mention the Amazon in 15.5% of their speeches.⁴ However, we observe four peak years (one standard deviation from the mean) in the figure: 1989, 1992, 2009, and 2019. The longitudinal trends in Amazonian speeches reveal a dynamic that model 1 does not show: peak years coincide with high-profile environmental events that gain large repercussions within and outside of the country.

Figure 2. Amazonian speeches in time.

In 1989 the Amazon appeared in 32% of all speeches. This coincides with the brutal murder of the environmental activist Chico Mendes in the last days of 1988. The incident caught unprecedented transnational media attention and then President Sarney (1985–1989) responded to this with a set of policies to address deforestation (Capobianco 2021). One of the responses was to host the 1992 Earth Summit (Keck and Sikkink 1998), which also coincided with the 1992 peak. During the summit, the Brazilian government announced the first large transnational partnership for the Amazon, the Pilot Programme of the G7 for the Protection of Rainforest, which brought a high number of financial resources to the region for public policy implementation (Hochstetler and Keck 2007).

The next peak is 2009, which coincides with the Copenhagen Summit. With deforestation rates decreasing since 2005, Lula led the delegation to the summit with a strong climate leadership self-image of ‘we deliver’ (Franchini and Viola 2019). From 2010 to 2016, except for 2015 when the Paris summit took place, we see a general decrease in mentions of the Amazon in all official presidential speeches. In this period, the urgency of environmental problems was increasing as deforestation rates were rising. The period was also marked by political and economic instability which culminated in the impeachment of President Rousseff in 2016, when mentions of the Amazon were at their lowest share.

We subsequently observe a steady increase reaching 24% in 2019, the first year of Bolsonaro’s presidency. At the time, international and national media brought attention to record burning of the Amazon, which led up to a red sky afternoon in São Paulo in August 2019. Right after taking office, Bolsonaro went on to revoked Brazil’s hosting status for COP25. Alongside this, Bolsonaro’s dismantling of environmental governance (Capelari et al. 2023) and the threats to leave the Paris Agreement had strong media attention.

Figure 2 also shows that the frequency at which the Amazon is mentioned, as a general topic in speeches, does not coincide with proposed environmental policy solutions. In 2000, for instance, Brazil adopted a strong legislation regulating the creation of protected areas (Silva-Muller and Faul 2022). In 2004, President Lula decreed the creation of Brazil’s Plan to Combat and Control Deforestation in the Brazilian Amazon. In 2008, Brazil announced the Amazon Fund, the largest fund of this kind ever created. Finally, in 2012, the revision of the Forest Code, debated in 2010 and 2011, pardoned thousands of illegal deforesters in the Amazon (Sant’anna and Costa 2021). Yet, these environmental policy landmarks do not seem influenced systematically when the Amazon as a general object of policy appears in presidential speeches.

4.3. Amazonian problem-constructions over time

Figure 3, below, illustrates the annual share of pure-types problem-constructions since 1985. Peak years identified in Figure 2 do not coincide with increases of specific problem-constructions in Figure 3.⁵

Figure 3. Pure-types constructions in time.

Economic integration problem-constructions were the most frequent from 1985 to 2009. This is especially pertinent during Cardoso’s presidencies (1995–2002), when construction of the Amazon as an issue of economic integration increased while environmental conservation constructions decreased. When Lula took office (2003–2010), constructions of the Amazon as problems of environmental conservation and social development increased. From around 2010 until the mid-2010s, especially during the Rousseff administration (2011–2016), environmental conservation and social development problem-constructions surpassed economic integration. However, with the presidencies of Temer (2016–2018) and Bolsonaro (2019–2022), economic integration constructions become the most favored Amazonian problem-construction again. We also identify a steady increase of sovereignty as salient problem-constructions from 2008 onwards.

While environmental policies did not coincide with peak years in general mentions to the Amazon (Figure 2), the trends identified above suggest problem-constructions might relate to policy. There is ample evidence that ties Lula’s environmental policies to the unprecedented decrease in deforestation between 2004 and 2012 (Arima et al. 2014, Assunção et al. 2015). This was also accompanied by a period of strong economic growth and large-scale social policies until the mid-2010s. The fall of economic integration and the rise of both social development and environmental conservation problem-constructions suggest an unprecedented balance between granting local livelihoods their social rights, environmental protection, and economic development.

In turn, policy developments taking place in the late 2000s might help us understand the steady rise of national sovereignty problem-constructions starting in 2008. The revision of the Forest Code, adopted in 2012, inaugurated a new phase of stronger representation of agri-businesses in government and congress and, consequently, strong civil society criticism of the government (Horochovski et al. 2016). In a period when sovereignty-related issues broadened from nation-states to a wider set of actors, presidents might become more assertive about Brazil’s sovereignty over the Amazon while rebating broad criticism from civil society actors with accusations of internationalization attempts. Therefore, the political forces in Brazilian politics that drove policy and discursive changes in problem-construction were long in the making. Bolsonaro’s problem-constructions, characterized by a unique combination of economic integration and sovereignty, is the strongest form of this shift.

Our operationalization of problem-constructions also foresees the possibility of presidents mixing multiple constructions within the same Amazonian Statement. Figure 4, below, shows that constructing the Amazon as multiple issues generally increased from 18%, in 1985, to 29%, in 2022, as a share of all constructions. Mixed-type problem-constructions in discourse offer more complex and multifaceted understandings of the Amazon as a problem. The increase in mixed-type constructions suggests that later presidents see the Amazon as an important mix of environmental, economic, social, and sovereignty-related problems. Indeed, we observe an increase in policies that attempt to address environmental, economic, and social problems, such as payment for ecosystem services conditioned on low-income residing in the Amazon (Silva-Muller 2022). Relatedly, this is in line with transnational processes such as the Millennium Development Goals and the Sustainable Development goals, that highlight the importance of human well-being in the economic, social, and environmental spheres.

Figure 4. Mixed-types problem-constructions in time.

We interpret more balanced use of all pure-types and the general increase in mixed-types Amazonian problem-construction as evidence that later presidents reconcile the binary between economic integration and the environmental conservation constructions that prevailed in earlier decades (1990 and 2000s). We see this as evidence of the Amazon becoming increasingly recognized as an intricate policy object in transnational politics.

5. Conclusion

This paper investigated how the Amazon has been constructed as a political problem over time and across spaces in Brazilian presidential speeches. We have four main findings. First, when presidents are far away from the Amazon region, they are more likely to construct the Amazon as an issue of environmental conservation. Within the Amazonian region, presidents usually construct the Amazon as an issue of social development and economic integration. Second, we find that presidents are more likely to construct the Amazon as an environmental conservation problem when environmental related issues are not urgent (e.g. decreasing deforestation rates). When environmental related problems become more urgent (e.g. increasing deforestation rates), presidents tend to side-step the issue and construct the Amazon as an issue of economic integration. Third, we find that high-profile environmental events, which receive strong international media attention, drive up the frequency at which the Amazon appears in discourses, but does not affect systematically how the Amazon is constructed as a problem. Finally, we find that presidents are increasingly more likely to mix problem-constructions and construct the Amazon as a complex and multifaceted object of policy.

Overall, our findings suggest that we must move beyond assuming the Amazon is a single issue to identify how both discourse and policy reflects its environmental, social, economic, and sovereignty-related complexity. In so doing, we can better understand the cohesive, the negotiated, and the contrasting problem-constructions forwarded by transnational actors in distinct policymaking instances. By capturing this variation in discourse, we can better understand why, when, and how, various overlapping policies directed at the same object are developed. Concretely, this implies that even as the environmental crisis becomes more pressing, we might not see coherent and cohesive environmental, economic, and social discourses or policies towards the Brazilian Amazon.

In terms of future research, we see at least three possibilities. First, scholars should investigate comparatively how problem-constructions change across settings and over time for other Amazonian countries or other policy objects. For example, recent literature demonstrates that reductions in deforestation in the Amazon lead to increased deforestation in the Cerrado biome (Villoria et al. 2022). By comparing the Amazonian statements and Cerrado statements, we could identify to what extent other problems are ignored in favor of the ‘Amazon problem’. Second, social media can change political communications in fundamental ways. The spread of political communications in social media could, on the one hand, contribute to diminishing the importance of physical settings and make presidents less likely to afford diverging problem-constructions. On the other hand, micro-targeting of social media algorithms could enhance how different problem-constructions are distributed to audiences. We look forward to works that investigate variation in problem-constructions given the advent of social media. Finally, our findings have implications for the scholarship on transnationalism. Scholars should investigate whether other transnational actors, as advocacy networks, have identified and acted upon contradictory problem constructions across settings. It could be, for example, that distinct policy making instances became closer over time, because advocacy networks moved across settings to highlight contradictions. In other words, if perceptions of audiences across settings explain problem-constructions, the mechanisms through which these perceptions diverge and converge should be explored.

Acknowledgments

The authors are extremely grateful to Graziella Moraes Silva, James Hollway, and Matias López for their support. The authors would also like to thank Anna, Federico, Mario, Rodrigo, and the two anonymous reviewers for their invaluable help.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/09644016.2023.2220639.

Notes

1. We also train the model to identify false positive matches in Amazonian statements. These are statements that contain the stem ‘amazon’ but are not meaningful as they might salute the governor of the Amazon, for example.

2. We conducted more direct tests of these effects, results are in the supplementary materials. While table 3 shows the limited effect of presidents’ party ideology; figure 5 confirms presidents have a similar share of speeches by setting, weakening the possibility of presidents picking settings according to their agendas.

3. All replication materials are available upon request.

4. For comparison purposes, the averages for other policy objects in the same corpus of speeches are inequality (13.7%), criminality (17.3%), inflation (19.1%), and unemployment (13.3%). This suggests the Amazon is a topic of relative high importance for presidents generally.

5. We run lag analysis to uncover hidden patterns and find no significant autocorrelations between peak years and problem-constructions (see Figure 6 in the supplementary materials).