Policy Dynamics and Government Attention over Welfare Policies: An Analysis of the Brazilian Case *

The recognition of issues as public problems and the ways governments prioritize them constitute focal points in the study of policy change and policy dynamics. In Brazil and other Latin American countries, social welfare systems and related policies have undergone transformations throughout the recent democratic period. This article aims to understand changes in the Brazilian social welfare agenda by means of an analysis of the attention given to social welfare policies at the federal level. contribution of this article is its use of the research strategy developed under the Comparative Agendas Project (CAP) to analyze the Brazilian situation. We drew on a set of unpublished datasets on the attention given by governments to social welfare policies from 1988 to 2018 that involves more than one thousand observations across six different datasets. The analyses are made at two different levels: first, we seek to understand macro trends and moments of continuity and inflection in social welfare policy by federal government administration. Second, we analyze the composition of the attention given to social welfare policies, thereby identifying the themes given the highest priority.

he ways in which social rights are translated into public policies vary historically and institutionally according to economic, cultural and socio-demographic factors as well as political choices. This gives rise to different 'worlds of social protection', according to Gøsta Esping-Andresen(1991). Protection systems that aim to guarantee social welfare also generate stratification in regimes where coverage and generosity vary. In Latin America, the division between insiders and outsiders has been a hallmark of social protection systems since their inception. The main goal of this paper is to apply an innovative methodology that allows us to understand changes in the attention given by government to welfare policies in Brazil in the period since re-democratization . This new analytical strategy, through the application of an international and comparative project that is recognized worldwide in policy process studies, helps us understand moments of continuity and discontinuity in the agenda of this important policy area.
There have been many studies that aimed to understand and explain the origins and patterns of the Brazilian social protection system, including classical historical studies -such as those undertaken by Wanderley Guilherme dos Santos (1979), José Murilo de Carvalho (2005) and Sônia Draibe (1993) -and those focused on the new social policy patterns that have developed since re-democratization in the mid-1980s (ALMEIDA, 2005;ARRETCHE, 2018ARRETCHE, , 2012ARRETCHE, , 2009ARRETCHE, , 2002. In this paper we do not intend to provide a new causal explanation for the origins and developments of the Brazilian social protection system since the Federal Constitution of 1988; our goal is more modest and complementary: we aim to shed new light on the debate by using an innovative methodological approach that can help in understanding (dis)continuities in the federal agenda.
Our research provides answers such questions as: that lead to the inclusion or exclusion of welfare policies from policy priorities?
Among the different alternatives available to deal with problems related to such welfare matters as food security, income, poverty, superannuation and disability, is it possible to find changes in government understandings and actions? When we analyze the different indicators of governmental attention in respect of the symbolic and decisional agendas, is it possible to find similarities in the levels of attention given to welfare policies? This is not the first time that such theoretical and methodological tools have been used to analyze Brazil's agenda-setting and policy-change processes (BRASIL, 2017;BRASIL and CAPELLA, 2019a;BRASIL and JONES, 2020;SUDANO, 2018). It is, however, the first time that a large amount of data has been used for the specific analysis of Brazilian welfare policies over a long time period. When we look at studies carried out by the international community, our research design stands out because the focus here is on multiple datasets that focus on a single policy area (social welfare policies). Traditionally, studies produced by researchers using the same methods and approaches (policy attention indicators over time used to identify priorities and changes in public policies) have favored longitudinal analyses with a single dataset across policy areas that have shown levels of attention given to the various policy areas and then considered the topics that stand out over time (BORGHETTO, CARAMMIA and RUSSO, 2018;BORGHETTO, SANTANA-PEREIRA and FREIRE, 2020;CHAQUÉS-BONAFONT, PALAU, and BAUMGARTNER, 2015; the constitutional provisions enacted in 1988. "It required the entire set of subsequent constitutional legislation, a process in which the ruling party played a critical role" (SÁTYRO and CUNHA, 2018). In this sense, the use of detailed indicators of governmental attention, internationally recognized but heretofore not applied to the Brazilian case, can make important contributions to the debate.
As a complement to the work of Marta Arretche (2018), which analyzes access to education, health and social welfare policies since the 1988 Constitution, the contribution of our research design is to produce an analysis based on the literature on agenda-setting, thereby filling a gap in the debate in the Brazilian literature by analyzing the relationship between policy attention and policy change. Our design, methods and theories have allowed us to identify continuities and discontinuities over time. Nonetheless, discussion of the main factors that explain these processes is beyond the scope of this article.
In this case study we look at the main trends in social welfare and development policies, including income transfer and food security programs, as well as programs and services aimed at specific social groups, such as children, young adults and the elderly. As our analyses show, the understanding of the importance of social policies has changed as have alternative solutions and the foci of such policies.
Building on the literature on agenda-setting and the role of policy attention in policy dynamics BAUMGARTNER , 2012, 1993;BAUMGARTNER, JONES and WILKERSON, 2011;BEVAN and JENNINGS, 2014;CHAQUÉS-BONAFONT, PALAU and BAUMGARTNER, 2015;MORTENSEN et al., 2011), the main analytical and methodological contribution of this article is its use of the research strategy developed under the Comparative Agendas Project (CAP) for Brazilian policy research (ANDRADE, BRASIL and CAPELLA, 2021;BRASIL, 2017;CAPELLA, 2019a, 2019b;BRASIL, CAPELLA and FAGAN, 2020;CAPELLA, BRASIL and SUDANO, 2015). This contribution accompanies other similar works that have sought to understand agenda-setting and changes in priorities vis-à-vis various policy areas in Brazil. In a specific study on health policies, Brasil and Capella (2019b)  There is no consensual definition of 'social policy' or 'social protection'. As we have learned since Marshall (1967), notions of social law are historically and socially situated depending on the concept of socially shared citizenship and thus have different political-social institutional arrangements. According to Kerstenetzky, the welfare state can be defined as "a set of public interventions aimed at promoting welfare and involving (some degree of) income redistribution" (KERSTENETZKY, 2012, p. 447).
For this paper we selected six different datasets to analyze social policy initiatives -such as social welfare, which was established as a policy by the 1988 Constitution, anti-poverty policies (such as cash transfer programs as the 'Bolsa Família' program (PBF) and the 'Benefício de Prestação Continuada' (BPC)) and policies that are aimed at specific groups, such as youth, women, children and the black population (JACCOUD, BICHIR and MESQUITA, 2017). In line with recent studies that have demonstrated the importance of attention at certain political moments and of windows of opportunity at the federal level for the consolidation of social welfare as a public policy, this article analyzes the level of attention given to welfare policies at the federal level by means of an innovative methodological strategy, mapping and analyzing the percentage of government attention given to welfare policies over time through different official documents: investiture speeches, addresses to Congress, provisional presidential decrees, presidential decrees, approved ordinary laws and approved constitutional amendments.
The article is structured in five parts, besides this introduction. The first section presents the theoretical framework of agenda-setting and policy dynamics.
The second, methodological section presents the data selected for this study and the tools adopted for the construction of databases on Brazilian government attention.
The third section is dedicated to the history of social policy development in Brazil. The fourth section presents an analysis and a discussion. The fifth section summarizes the main findings and indicates possible advances in this research agenda.

Agenda-setting, policy attention and policy dynamics
Over the past thirty years, theoretical and methodological contributions to studies of the agenda-setting process have transformed the literature and the ways in which we understand the process of prioritization and definition as well as the dynamics of public policies. These models (Multiple Stream; Punctuated Equilibrium Theory; Advocacy Coalition Framework) come from the US literature of others. The levels of attention given to these issues indicate consequent changes in public policies, be they incremental or of substantial impact (punctuations).
Punctuated equilibrium is a theory of policy dynamics that details a set of mechanisms that lead to policy change. Policy change is caused by changes in the preferences of policymakers that do or do not occur in response to different situations affecting political, social and economic systems as well as the attention given to certain matters and the relevant information processes. Unlike institutional-based models that explain policy changes, such as the election- In the following excerpt, Baumgartner, Jones and Wilkerson (2011) summarize this innovation and the importance of the findings set out in their General Punctuation Thesis. They point out the relationship between attention, agenda-setting and policy change: The typical agenda-setting study from the policy literature suggests a highly dynamic policy-making process where changes can occur more rapidly than a preference-based approach would lead us to expect. Changes in information are central to any explanation. The scarcity of the attention of policymaking institutions is critical. There are many more problems than governments can possibly attend to, and each problem may be extremely complex. These challenges lead to attention scarcity. For example, discussions of poverty may be focused on the severity of the problem or on the efficiency of the solutions but rarely both at once. Both are relevant dimensions, but attention typically is not divided in proportion to any comprehensive assessment of the relative weight of the diverse elements of an issue. (...) The policy agenda-setting perspective also highlights the multidimensional nature of policy process -how attention scarcity can cause policy-making attention to shift from energy to health care to immigration to the economy to war, climate change, or human rights. (BAUMGARTNER, JONES and WILKERSON, 2011, p. 951). The concept of institutional friction is central to the PET approach and to explain policy incrementalism and change. This concept is present in the literature on organization theory, sociology and behavioral psychology. The concept of friction is related to organizational patterns and culture (symbols and human behaviors) that generate stability and define the forms of operation that tend to preserve the status quo rather than continuously promoting change. In addition to being characterized by organizational and cognitive friction, institutional friction is governed by institutional rules. These are pre-established rules of action that restrict political action or impose mandatory processes to be followed by decisionmakers. These rules, such as those set out in a country's constitution or the procedures governing the approval of new legislation, restrict or at least limit major changes so that they cannot occur frequently and easily.
The stability imposed by the two kinds of friction, cognitive/organizational friction, and institutional friction, does not cause universal gridlock, with a system awaiting elections to point to change. However, it is a retarding force that interferes with the smooth adjustment of a political system to changing information signals from the policymaking environment. Change occurs only when the informational signals from the external world either are extraordinarily strong, on the one hand, or when the signals accumulate over time to overcome the friction. As a consequence, policy -making systems remain stable until the signals from outside exceed a threshold, and then they lurch forward -that is, a policy punctuation occurs; afterward, they resume 'equilibrium' (JONES and BAUMGARTNER, 2012, p. 08).
Institutional friction is variable. Different institutions, actors and venues exhibit different levels of friction. In Brazil, some indicators of governmental attention may be subject to a greater or lesser degree to institutional friction. For instance, the symbolic agenda that involves presidential speeches, such as a president's inauguration speech and annual speeches at the early stages of the policy-making process, are not debated at committee level and do not require that the legislative and the executive branches come to an understanding.
The topics addressed in such speeches reflect a broader agenda, characterized by 01. an absence of restrictions regarding content (i.e. any topic can be addressed); 02. an absence of decision costs; and 03. low institutional friction. In general, the symbolic agenda is related to the first steps in the policymaking process, in which institutional costs are lower than in later steps (BAUMGARTNER and JONES, 2005, p. 175). In contrast to the symbolic agenda, government indicators such as the 'approved ordinary laws', which represent a decisional agenda (the inner circle of the decision-making process), demand a competitive process from proposal to approval (debated in committees, voted on by policymakers, tested constitutionally where called for). Obviously, these indicators are subject to more institutional friction than are symbolic agenda speeches. Throughout our analysis, we will show how the degrees of attention given to social policies can vary according to the different indicators used. The concept of institutional friction will be widely used in the analysis to understand increases and decreases in the levels of attention given to a given policy, as well as the relationship between the types of indicators selected.

Data and methods: the comparative agendas project and the Brazilian case
The Legislative production is also analyzed as an indicator of attention, both in respect of the executive (presidential provisional decrees, executive orders) and the legislative branches (approved ordinary laws and approved constitutional amendments).
The data selected for this research present peculiarities in respect of the legislative process and, therefore, institutional friction, as noted by Jones (2005, 2002). Documents such as addresses to the National to approvals in a competitive process of defining priorities, these data (i.e., approved laws) become an increasingly accurate gauge of policy change.
The variety of indicators selected by this study allowed us to measure and analyze different approaches in respect of social policy dynamics in Brazil.
From the symbolic agenda -more fluid and with fewer institutional barriers -to the legislative agenda -a source of indicators of attention that have gone through such processes as that required for laws to be approved -this study presents indicators and results and compares them over time in order to demonstrate the relationships between attention levels vis-à-vis the selected dataset.
This study brings together qualitative and quantitative perspectives involving multiple research tools and methods. We used an information coding system to create a database that synthesizes and organizes information related to the attention given by actors to sectoral policies. Among the main actors involved, the presidency, the legislature and public opinion are considered central to this type of analysis. We used a method established for the first time in the US Policy Agendas Project (BAUMGARTNER, GREEN-PEDERSEN and JONES, 2006), in which public policy inputs and outputs are ranked using a standardized coding system to produce valid comparisons within nationally-specific and international databases.
To map attention given to sectoral public policies over time, the research design proposed by the US Policy Agendas Project and replicated in this study builds databases on governmental attention using a content analysis approach, with the transformation of raw materials, such as laws, texts and speeches, into major topics and subtopics as a methodological tool. The goal of this process is to comprehensively code large datasets to focus on the attention given to a specific policy issue over time. The coding process follows a common set of rules in order to accurately code the policy focus of each observation. Attention is mapped according to the substantive policy of each item (each ordinary law or each provisional presidential decree) and not according to the target audience or arena where it is defined. This means that the analysis of the selected item is interpretative and seeks to identify in which sector of public policy (only one topic) each observation can be identified according to a list of topics (for the code book see Table 01).
We seek to analyze the macro sector into which the public policy information under examination is inserted, this being selected from a predefined list of sectoral policies called 'major topics'. To ensure comparability between the various cases we consider, these major topics are fixed, i.e., they are not modified according to the specificities of each country's public policies. In other words, it is assumed that national variations can be grouped into these pre-established major topics.
There are some challenges arising from the use of this coding system, especially with information that is sometimes intersectoral and cross-cutting and which is coordinated by different ministries or agencies. It is not possible, according to the coding rules, to classify the same information into two distinct variables (subtopics). This means the classification process is of utmost importance 1 .
______________________________________________________________________________________________ 1 All observations are manually coded twice by trained research assistants. Dual coding serves as a reliability and credibility mechanism for coded observations. When coders disagree, a third researcher reads the coders' remarks and resolves the deadlock by means of a final decision. This standardized coding and database-building process means that raw information such as speeches, official texts and laws can all be uniformly and comparably transformed into major topics and subtopics. When coding selected observations, researchers follow a common set of rules to accurately code the policy focus of each observation. Table 01 shows the list of major topics used by all CAP member projects. In addition to the major codes, which set the different sectoral policies, the coding system also classifies observations into more specific subtopics, i.e., 221 subtopics that are hierarchically below and necessarily linked to each major topic.
Thus, for each major topic, there is a variety of smaller, more specific subjects inserted within the major policy theme. For example, under the topic of 'energy' subtopics represent different types of energy policies, such as wind, thermoelectric, renewable and oil, among others. Scholars can use these data to perform analyses at higher or lower levels of aggregation.
The definition of our major topic of interest, i.e., 'welfare policies', does not include education, health, employment or income generation policies. According to the definition adopted by the group, welfare policies refer to assistance, social To avoid confusion, we wish to point out that the method used in this study does not take into account the importance, impact or any evaluative aspect of public policies that appear in the documents we analyzed. Our method highlights the appearance of topics in public policies over time. Thus, a law that creates a national social assistance system and a law that regulates philanthropy, for example, are codified equally according to the number of times that they appear.
Therefore, qualitative analyzes, as institutional and historical approaches, supported by the literature on welfare policies, interpret these data and identify special or high-impact policies.
In this paper, we define two distinct levels of social policy attention analysis in Brazil: one that takes into account the attention of the general social policy in comparison to the other sectoral policies proposed in Table 01 and another that disaggregates the general attention given to social policy into smaller and more detailed issues within this sectoral policy, while concentrating the distribution of social policy subtopics, as shown in Table 02.
After the systematization of the data, one of the main challenges faced by studies on agenda setting and policy change is to determine which information, actors and arenas are important when analyzing attention to social policies in Brazil, as discussed in Jaccoud, Bichir and Mesquita (2017). Following the recent literature, especially the work already developed by the CAP groups, this article uses the data produced by the Brazilian project. There are six types of data from two different arenas: the executive branch, which involves speeches made by the president, addresses to Congress, provisional presidential decrees and presidential decrees; and the legislative branch, with information on approved ordinary laws and constitutional amendments. To ensure comparability with related projects developed internationally, neither ordinances, resolutions, instructions or normative resolutions were considered in the corpus that we analyzed 2 . through a fragmentation process, whereby 'sentences' are created according to the appearance of a new policy issue. See an example in the following passage from an Inauguration Speech: "It is absolutely necessary for the country to grow again by generating jobs and distributing income to those in the greatest need". This text is separated into three different 'sentences', since the same sentence refers to three different sectoral policies. The sentences are: 01. "It is absolutely necessary for the country to grow again" which identifies a concern about macroeconomic policy; 02. "generating jobs", which draws attention to employment policies and, finally, 03. "distributing income to those in the greatest need" which reflects a concern about policies to combat social and economic inequalities. As can be seen in the example above, the construction of the sentence reflects the way in which attention is given to different policy areas. After the fragmentation of the text into 'sentences', each 'sentence' is allocated to a major topic and a subtopic. Texts of this type correspond to hundreds of 'sentences'.
To guarantee credibility and methodological confidence, all legislation data and all sentences were classified twice, by two different researchers, according to CAP methodological specifications, and as detailed by the group that produced the codification of all data. It is important to note that the selected data from major topic 13 (welfare policies) used in this research, went through a new check, carried out specifically for the development of this research, where the authors analyzed all the codified data ratifying the codification performed. The results are shown below in Table 03. Table 03 presents a summary of the database indicating the amount of data analyzed from each sector.  According to Almeida (1995), the area of social assistance was less changed by this reform cycle than such areas as health, housing and education. Almeida In spite of the fact that inequality is a well-known characteristic of Brazilian society, social assistance policies have not been always central to the federal agenda.
Recent studies, especially Arretche (2018), have shown that policy decisions are central to understanding the process of inclusion of marginalized groups that began with re-democratization. This process has been marked by advances and setbacks in a to and fro between politics and policies that needs to be analyzed in detail, in addition to approaches that are based on historical-institutional arguments. In the next section we contribute to this debate by showing, in a nuanced and empiricallybased way, which federal administrations devoted attention to social protection and what the main themes on their agendas were. In the section that follows it, we will resume the characterization, for use in the analysis of the empirical data, of the main continuities and discontinuities across federal administrations.

Analysis
Two levels of analysis will be described in this section, the major topic level and the subtopic level, as presented earlier. Firstly, we will compare the attention given to welfare policies to that given to the other policy areas that are classified as major topics. Thus, we will consider the percentage of attention given to major topic 13, i.e., welfare policies, relative to the other major topics listed in Table 01. Such an analytical strategy allows us to compare the distribution of governmental attention to policy areas over time.
The following graphs (01 and  Temer, assumed the presidency for the remaining two years of each term of office. We start the analysis of the data by presenting an overview of the main topics that appear as priorities in presidential inauguration speeches Although a diffuse sense of 'social debt bailout' was present in the context of redemocratization, as noted by several authors (ALMEIDA, 1995;ARRETCHE, 2018;DRAIBE, 1993), the initial foci were macroeconomic stabilization and inflation control (so-called first-generation reforms) as a precondition for any expansion of social investment, and the reorganization of government, civil rights, liberties, democracy and the public administration (ANDRADE, BRASIL and CAPELLA, 2021).   As we will see in the following tables, welfare policies were given little or no attention in inaugural speeches in the early years of democratization (from 1985 to 1990). Neither were social issues presented as a priority in Lula's first and second terms, welfare policies accounted for 20% of the topics given attention. Incrementalism is clearly apparent at certain moments. The first, between 1990 and 2002, includes Franco's and Cardoso's first and second terms, as well as Rousseff's first term. Incrementalism is inferred when the data do not exhibit large variations or peaks in terms of attention but maintain a stable average across time. Specific moments are treated as snapshots and represent the level of attention being given to a specific issue at that moment.
By increasing or reducing the number of such snapshots, peaks may appear or legislative productions from the legislative branch (constitutional amendments and ordinary laws).
These indicators form a dataset that reflects a different type of agenda than the symbolic one. This dataset reflects a more restricted agenda that is subject to greater institutional friction as it has perforce gone through long debate and prioritization processes. Compared to the high levels of attention given to welfare policies in the symbolic agenda, the institutional friction produced by the legislative process potentially reduces the levels of success achieved by and attention given to welfare in approved legislation. The comparison is relevant because it shows that the levels of attention given in speeches cannot always be sustained sufficiently for intentions to be transformed into legislation.
From the Graphs 03 and 04 one can draw two major conclusions based on the agenda-setting and policy change literature. The first major conclusion is directly related to the barriers and limits imposed by environments with high institutional friction such as that suffered by legislation that is necessarily the subject of debate and approval in multiple venues. In the presidential speeches presented in Graphs 01 and 02, the average attention given to welfare policies was approximately 10% to 12%, and it peaked at 14%. This denotes a high priority being given by presidents to social issues in environments with low institutional friction via statements characterized by the expression of promises and intentions. By contrast, the welfare legislation that ended up being approved in the last 30 years, which had to make it through the grueling approval process, accounted for a mere 1.5 % of the total (See Table 03). We can therefore conclude that there is a mismatch between the attention given to welfare in the symbolic agenda and the resulting attention given to welfare in approved legislation.
Our second major conclusion is related to the process of incremental change that is characterized by stability, continuity and small changes in public policy.
When analyzing the data from 1988 to 2018, we observed a stable line without large ups or downs. Especially after the 1990s, welfare policies clearly play a discreet role in which they are always present but gain little attention in the form of action being taken by the presidency or the legislature. We can therefore conclude that welfare is ever-present but not a matter of great priority. Before moving on to a more detailed analysis of the subtopics, it is important to note that the design of this study does not consider the importance, impact or any evaluative aspect of public policies created over time.
Especially in this first level of analysis, this research design demonstrates, with 0,00% 0,50% methodological rigor, the level of attention given by selected actors to specific subject matter. Our first analysis found that welfare policies were given less attention during the high-friction legislative process than in low-friction discursive contexts.
Our subsequent analysis considers the subtopics presented in Table 02, and allows us to verify the differences and similarities between governments in terms of the provision of services, themes and target beneficiaries. One of the most interesting results of the analysis is presented in the following chart. Two types of indicators were selected from within presidential decrees: anti-poverty policy indicators (subtopic 1302) and policy indicators promoted through volunteering, partnerships and recognition of charities (subtopic 1305).
Graph 08 summarizes two central patterns of dispute in the field of social assistance. The first pattern relates to disputes over how services should be provided, i.e., whether they should be provided directly from the State or indirectly (JACCOUD, BICHIR and MESQUITA, 2017); the second pattern relates to the placement of anti-poverty policies, particularly income transfer programs, in the context of social assistance (BICHIR, 2011). As other recent analyses have found (JACCOUD, BICHIR and MESQUITA, 2017), we observed a downward trend in terms of governmental attention given to policies that fight poverty, especially in the context of the Temer administration.

Discussion
The quantitative analysis at the major topic level indicated that, in general, inauguration speeches and addresses to the National Congress mention themes in the field of social assistance less than other subjects, such as macroeconomics, the organization of government and public administration. Important exceptions are the second Cardoso government and the Lula government, when the fight against poverty and the institutionalization of social assistance policy were given more emphasis. These results are in line with the literature, as the early years of redemocratization were marked by attempts to combat inflation, increase economic stabilization and enhance debt control, as well as the organization of public administration, these being the so-called first-generation reforms (MELO, 2005).
From the 1990s onwards, the so-called second-generation reforms focusing on social policies have become more prominent. Transformations in the field of social assistance occurred late and have really only achieved much impetus since the early 2000s, when the SUAS began to gain institutionality and advance the construction of State capacities in the provision of social assistance services and conditional cash transfer programs, especially the PBF. With the growth of the fiscal deficit from the second Rousseff administration onwards, macroeconomic concerns have returned to the agenda, and the penetration of neoliberal ideas has led to a sharp retrenchment in terms of social assistance and social development issues toward the end of our period.
Our analysis of the legislative production of the executive and legislative branches has indicated two central aspects: 01. approved legislation is more relevant than speeches, in spite of the striking difference in terms of the attention given to social welfare according to the presence of institutional friction; and 02.
logical incrementalism remains the name of the game, as it has since Lindblom, i.e., change processes are often incremental and any analysis of long-term government attention reveals more continuity than major transformations, especially when compared to the occasional findings of cyclical analysis. In the specific case of social assistance, this point has been very well made by Margarites (2019), who demonstrates by means of rigorous documentary and content analysis in a study of the 'Assembleia Nacional Constituinte', that the process of institutionalization of social assistance has long-lasting roots.
At the subtopic level, important nuances emerge between governments.
While the level of attention given to social assistance and equality promotion is generally low, there are clear differences between governments' substantive welfare policy agendas. Most 1988-2018 governments focused the attention they gave to the area to the relatively more general themes of social assistance and poverty alleviation, but two governments in particular gave attention to differentiated social welfare themes and agendas, with relevant legislative proposals and an influence on the overall trend; these being the Cardoso and Lula governments. Not coincidentally, several social policy analysts in Brazil have focused on the specific contributions of these two governments, particularly in the field of social assistance (ABRUCIO, 2005;ALMEIDA, 2004;BICHIR, 2016;FRANZESE and ABRUCIO, 2013;LÍCIO, 2012). The different ways that welfare policies are implemented can also be observed at the subtopic level: in the Cardoso government, policies to combat poverty were implemented primarily through civil society organizations rather than government; cash transfer programs were begun and expanded but in a fragmented and diffuse way. Under the Lula administration, the fight against poverty involved increasing the size of and the coordination between income transfer programs (especially with the creation of the PBF), through increased integration between benefits and services (BICHIR, 2016(BICHIR, , 2011 and the expansion of a social assistance network with direct State provision at its core. Thus, another maxim of public policy analysis has been reinforced, i.e., politics cannot be ignored in policy analysis, as social policy views, ideas and projects affect the ways institutional policy is made (JACCOUD, BICHIR and MESQUITA, 2017).

Final considerations
The aim of this article was to contribute to the debate on social welfare policies in Brazil and the debate on agenda-setting and change in government performance, based on an innovative analytical-methodological strategy that was developed internationally and applied to the case of Brazil.
The analysis considered the period from the promulgation of the Federal Constitution of 1988 to 2018, a period of seminal importance in the construction of social policies in Brazil. We drew on the literature and made an original analysis of an extensive documentary corpus, which we approached using quantitative content analysis strategies. We demonstrated that the institutional fragility afflicting social welfare in Brazil began to be dealt with after redemocratization. We found that social welfare did not lie at the heart of government concerns and has been dealt with in a variety of ways during our period.
We also sought to underscore the importance of policy analysis in general and changes in the federal government agenda in particular, especially when considering long periods of time, for observing the trends and inflections that can be hidden behind occasionally impressionistic conjunctural analyses. By working with different levels of aggregated information, we were able to look at both the field years. The analytical and methodological strategy that we employed reinforces the importance of considering social protection systems that go beyond dimensions such as social spending, as has been discussed in the specialized literature for some time.