Political Communication, Television Advertising and Elections in Brazil: A Longitudinal Analysis of Two Major Parties’ Use of Free Electoral Advertising Time

A longitudinal analysis of the use of televised electoral advertising by the two political parties that have been the central actors in presidential elections in post-democratization Brazil: the Workers’ Party (hereinafter referred to using the Portuguese acronym the ‘PT’) and the Brazilian Social Democracy Party (hereinafter referred to using the Portuguese acronym the ‘PSDB’). Our objective is to identify patterns and changes in Brazilian electoral advertising between 2002 and 2018 with reference to Brazil’s system of free electoral advertising time (hereinafter referred to using the Portuguese acronym ‘HGPE’), more of which has been allocated to these two parties than to any other. Despite the increasing use of other forms of electoral campaigning, HGPE is central to presidential contests and demarcates political time and parties ’ agendas. It is therefore relevant to understand how use of this campaign component has been developed over time, taking into account two factors: messages and strategies. Our quantitative content analysis considers 90 days of PT and PSDB block format advertising during the first rounds of five election contests. Our results reveal that changes in messaging tend to be decided on by the parties (with the PT emphasizing public policies and the PSDB emphasizing image building) whereas the strategies employed depend on the context of each election and are less specific to either party.

compare the elements of the parties' HGPE programs and their performance in television advertising in the five presidential contest that we looked at, by seeking to identify ruptures and continuities, both between the parties and between elections. We considered the total number of the parties' programs in the first rounds of the elections of 2002, 2006, 2010, 2014 and 2018 and focused on the two fundamental elements for characterizing HGPE use: the theme presented and strategies employed.
The article is divided into four parts. The first part discusses the roles assumed by the two parties in Brazilian presidential contests. The second part concerns the centrality and characteristics of HGPE as a campaign tool. The third part details our research, as well as the variables and categories analyzed. The fourth part presents our data analysis and conclusions.

Presidential elections in Brazil and the roles of the PSDB and the PT
This article analyzes the development of the use of HGPE by the only two political parties whose presidential candidates were elected to office in the two decades subsequent to the re-democratization period: the PSDB and the PT. This duopolization of the presidency over such a long period transformed Brazil's multiparty system into an essentially bipartisan one -at least insofar as the presidency is concerned -and had an important effect on local elections (LIMONGI and GUARNIERI, 2014). This led to other parties orbiting around the two poles of the PT and the PSDB (D' ARAÚJO, 2011) and made the two parties the most mentioned in public opinion surveys of party political sentiment (SILVA, 2012) the institutionalization of democracy after the end of the military dictatorship, and for being based in social democracy, understood as the defense of inclusive social reforms (D'ARAÚJO, 2011, p. 67). There was an ideological proximity between the two parties that was visible until 1992; from the 1994 elections onwards they began to diverge. This divergence has led to electoral polarization in respect of the best economic model for government (FREITAS, 2018, p. 550).
The PSDB stayed in government for eight years, starting with the election of President Fernando Henrique Cardoso (FHC) in 1994 and continuing with this reelection in 1998. The FHC government was marked by 'governability' that was made possible by Congressional support that allowed the Presidency to enact its agenda -so much so that the legislature at times distanced itself from public demands. Monetary stability and reforms to control inflation and incorporate Brazil into the global economy were prioritized, unlike socio-economic inequality and similar matters, which were not LIMONGI, and VALENTE, 1999).
At the end of its second term, the FHC government enjoyed the approval of only a minority of voters (DATAFOLHA, 2002). In 2002, the PT candidate Luiz Inácio Lula stood for reelection in 2006 and, according to Machado (2009), campaigned with a discourse of "defending the continuity of the economic growth model, while simultaneously enabling development in the social area" (MACHADO, 2009, p. 163 (BORGES and VIDIGAL, 2018). This combination of factors led to the election of Jair Bolsonaro (PSL), whose campaign was essentially centered on the figure of Bolsonaro himself and cannot be considered an indication of the strength of the PSL party (DIAS and FERNANDES, 2020), which did not enter into a coalition with any other parties in the election.
It is worth noting that in spite of their different ideological and administrative approaches, both PT and PSDB administrations were "characterized by financial stability, consistent social policies aimed at reducing social and regional inequality, a search for economic growth, pro-environmental policies and respect for legal and constitutional provisions" (D' ARAÚJO, 2011, p. 66). Over the period of these governments, economics became central to public and electoral debates. Brazil's history of inequality meant that social policies were also at the center of electoral debates (REIS, 2014), as reflected in HGPE.
With the benefit of hindsight, it is clear that voters tended to decide whom to vote for on the basis of public policies (DIAS, 2012). This can be observed geographically, whereby a given location's HDI, as a measure of economic, health and educational data, can serve as an indicator of electoral success. Twenty years of electoral contests between the PT and the PSDB showed th at "votes for the governing party are associated with dependency on public policies in given geographies, and votes for the opposition depend on the failure of these policies" (CERVI, 2016, p. 313). This is the reason why when the PT was in government it gained more votes in states with higher HDI growth, whereas the PSDB gained more votes in states with lower HDI growth (CERVI, 2016). This is of interest for our analysis in terms of the PT and the PSDB 's use over time of HGPE to focus on public policies. In the 2002 to 2018 elections, this means thinking about how the two parties leveraged public policy themes from their respective incumbent and challenger perspectives. As the tool we are looking at is HGPE, the next section will consider its main characteristics and the possibilities for analysis.

HGPE as a source of electoral information
Advertising is currently a relevant tool in electoral contests and can be considered a relevant variable in the voter decisions, especially where votes are cast on the basis of rational choice, given that this is more susceptible to media and campaigning because of the weight of the informational variable in decision making (TELLES and MUNDIM, 2015). This is especially true where there are undecided voters and in democratic scenarios with high electoral volatility. Election campaign advertising is defined as a space in which candidates and parties present themselves to the electorate, especially through the media, from their own space and not through the press. Party and candidate advertising is especially important because it allows for autonomy and control over messaging (HOLTZ-BACHA and KAID, 2006). However, as such campaigns are not watertight, there is a modernization process underway that directly affects the way electoral advertising is produced (MANCINI and SWANSON, 1996) and is associated with increased professionalization. first-past-the-post elections, it offers a space that allows for both thematic discussion and image formation, especially for coalition candidates with more HGPE time. In proportional elections, HGPE has the primary function of presenting the candidates so that the voters will search for more information about them by other means.
The first empirical studies of HGPE use investigated the discursive strategies employed by the candidates and the structure of their HGPE (ALBUQUERQUE, 1999;FIGUEIREDO et al. ,1997 HGPE has been obliged, therefore, to divide its space with an increasing plurality of other media, in addition to the other challenges television advertising faces.

HGPE themes and strategies analyzed over time
This article analyzes two specific variables over time, i.e. campaign themes and discursive strategies, which are looked at in terms of content and mode of transmission. Time can be a vector of changes that can be identified in the elections that we looked at and can shape different political interests longitudinally.
Observing the elections longitudinally makes sense considering the importance of time as a factor in shaping them. Among other studies, we find that Figueiredo et al. (1997) were a useful source for consideration of strategy and that the views of Dias' (2013; in respect of parties and their different approaches were a valuable source for consideration of HGPE. We wish to reinforce the fact that our study concerns neither the development of a campaign in the course of an election year nor evaluations of governments, as per Stimson (1976) with his cyclical model, but constitutes a longitudinal analysis that brings together different campaigns over almost 20 years.
We treat time as a unit of selection for comparisons and not as an explanatory variable.
Regarding the first dependent variable, our concern is to identify which themes and/or proposals are present in the electoral contest, which theme is related to a specific type of message and which theme seeks to identify problems and propose solutions (ALBUQUERQUE, 1999;ALBUQUERQUE and TAVARES, 2019). Although some studies have found that campaigns tend not to propose a great deal and to emphasize image formation to the detriment of public themes (SOUZA; CERVI, and SANTOS, 2009), others find that themes gain space among the types of messages transmitted (MASSUCHIN et al., 2016;PENDIUK, 2013) and that issues that refer to the context of the election can be of great prominence, as with economics in the 1994 election (PORTO and GUAZINA, 1999).
Considering the different approaches to use of GPE, our hypotheses are that (H1) the weight of thematic content differs by party throughout the trajectory of the five elections, with such changes as occur explainable by reference to the parties themselves and not by temporal evolution. This hypothesis is based on the understanding that political parties differ ideologically and therefore tend to engage in advertising differently, in the same way that they tend to form coalitions that match their ideological preferences (CARREIRÃO, 2006). Party specificities come first, despite differences inherent to the candidates themselves since, especially in the case of the PSDB 4 , candidates were replaced in one election after another, as were the positions occupied in each election. This is related to the party differences exposed in previous topics. The extent to which HGPE is used thematically varies according to the party distinctions and not any temporal evolution that might have changed the way that HGPE programming is constructed or brought the parties closer together.
Public themes compete for space with other kinds of HGPE content, such as image formation (ALBUQUERQUE, 1999). There is also the meta campaign, a type of auxiliary message with facts external to HGPE, such as research or the candidate's agenda, inter alia.
Discursive strategies are used with a view to persuading and convincing voters (FIGUEIREDO et al., 1997). Albuquerque and Tavares (2019) point out that HGPE already has pre-defined times and structures, which does not allow for radical production changes. The strategies are responsible, complementarily, for the way candidates choose to speak to the public, by highlighting different elements and ways of communicating content. These strategies have changed as campaigns have modernized. Dias (2013) shows that candidates can emphasize political parties, which is relevant for institutions. Tavares (2013) discusses how candidates can take advantage of political support from patrons in their advertisements, while Borba (2015) and Lourenço (2009) highlight the use of negative campaigning. There are also strategies that characterize the discourse of the incumbent president and/or his or her challenger (FIGUEIREDO et al., 1997). The strategies chosen are linked to the party's position in the contest rather than to the party itself.
In view of the foregoing, we also hypothesize (H2) that strategies vary according to a given candidate's position in an election, with no significant temporal or ______________________________________________________________________________________________ 4 In some cases the PT nominated the same candidates in multiple elections, unlike the PSDB. partisan influence. Party and time therefore have less weight in the definition/maintenance of a strategy, unlike HGPE theme-setting.
In respect of strategies, we focused on four elements -'endorsement, attack, call for change and use of office' -all of which are prominent in the literature and present in the scenario analysis of Figueiredo et al. (1997). They were less interested in the effects of these elements than in the ways they are used according to the electoral scenarios that distinguish competitors. 'Endorsement' and 'use of office' are typical of candidates whose party is in power while 'call for change' and 'use of office' are typical of challenger candidates.
Unlike Stimson's repeated cycle (1976), here we see perennial characteristics, such as thematic distinctions, as well as characteristics that change precisely because each contest takes place in a specific context. The PT and the PSDB were at times incumbents and at times challengers, which was an important factor in the strategies they adopted, according to the logic of the possible worlds and the evaluations made by Figueiredo et al. (1997).

Methodological aspects of the research
Following in the footsteps of Bauer (2012) and Krippendorff (1990), the CPOP- These were categorized according to elements present in the literature (ALBUQUERQUE, 1999;FIGUEIREDO et al., 1997) that have been maintained over time, which allowed for longitudinal analysis. The collection was made using a coding book containing several categorical variables that characterize the content of each program segment -i.e. an autonomous video excerpt -with thematic, speaker and defined environments PANKE and CERVI, 2011).
We had to demarcate our field, as reflected in our decision to focus on the PT and the PSDB, the parties that disputed the five elections, whether they obtained a higher percentage of votes or more time on television. The fact that the same parties participated in the five elections allowed us to identify aspects that were maintained ______________________________________________________________________________________________ over time. We chose to restrict our analysis to the first round of each election, which gave us more than 19 hours of video content in total.
Our analysis looks at two characteristics: strategies and themes. The strategies in question concern the candidate's communicative persuasive resources (FIGUEIREDO et al., 1997). The CPOP research group classified the strategy variable according to 10 categories, which can appear concurrently, and which are aggregated here into four types: 'endorsement, attack, call for change' and 'use of office'.
'Endorsement' refers to an appearance by another actor on an HGPE segment as a means of reinforcing the vote for a given candidacy. 'Attack' seeks to demobilize votes for an opponent (BORBA, 2015). 'Call for change' reinforces the message that 'it is time for a change' by highlighting the situation that needs to be changed (FIGUEIREDO et al., 1997). 'Use of office' encompasses messages in which the candidate addresses his or her political and administrative experience in his or her activities prior to the election. All are shown in Table 01. They include challenging strategies, which call for change and attack the status quo by emphasizing criticism of the 'bad' and presenting the viability of a new and 'good' government. The strategies of incumbent-aligned candidates tend to present the qualities of the current administration by characterizing the status quo as good and proposing improved continuity (FIGUEIREDO et al., 1997).
In this study we employed the thematic categories proposed by Albuquerque (1999). Themes matter for the construction of persuasive and strategic campaign communication that can address the incumbent government's proposed program, discuss issues of interest to the public, present candidates and coalition parties and address aspects of the campaign itself, such as any rules or specific events.
We started out with 34 mutually exclusive possibilities divided into three main groups, hours of video content 8 and was analyzed with reference to the variables described above. We employed an analytical approach so as to make feasible a longitudinal and comparative analysis. This analysis is described in the next part of the article and considers the proportional time invested by the parties in each election and the hypotheses set forth above: H1: the weight of thematic content differs by party throughout the trajectory of the five elections, with such changes as occur explainable by reference to the parties themselves and not by temporal evolution; H2: strategies vary according to a given candidate's position in an election (incumbent or challenger), with no significant temporal or partisan influence.

Content of HGPE messages: political-party distinctions in the use of advertising
The first characteristic we explored was HGPE content, which was divided according to the themes of public policy, image building and meta-campaigning.
Comparison the parties in Table 01 reveals that PT used its HGPE more thematically, while the PSDB focused its HGPE on image building, usually in respect of the candidate him/herself, as well as the images of his or her opponents, the party, voters, the country, and so on. There was therefore a noticeable difference between the parties that could be considered. The PT's public policy issues percentage exceeds that of the PSDB by 12 points.
In respect of time dedicated to image building, the PSDB exceeds the PT by 13.7 percentage points. There is thus an inversion of content priority. In the case of metacampaigning, the two parties presented very similar percentages, with a difference of less than one percentage point (i.e., 0.3) between the PSDB and the PT. The comparison shows that the two parties are primarily distinguished by an inversion of theme and image. The PSDB expended similar amounts of HGPE on the two main content types, whereas the PT lavished palpably more space on campaign themes than image building. Table 02 details the percentage of party time dedicated to each content in each election longitudinally, in addition to the difference in percentage points for the average. Temporal differences reinforce that fact that the two parties present themselves in a clearly distinct manner according to the weight given to theme and image building in their HGPE, even when the complex relationship between the political and media spheres is taken into account (MARQUES and CARNEIRO, 2018).
In the case of the PT, its 2002 and 2018 meta-campaigning HGPE is longer than the average, as is its image building in 2006 and 2010, and its public policy HGPE in  To complement the analysis of HGPE content, the Table 03    Regarding the specific themes observed in this first part of the article, the data indicate that television advertising is still a space for the thematization of the first round of an election, however this is not its standardized function, nor is it evenly Thematization of HGPE is proportionally more used by the PT than the PSDB. The PSDB makes greater use of image-building and shows more non-linear change over time. We cannot say therefore that there is a temporal tendency towards thematization in respect of both parties. This can happen, but in this case only did for the PT.
This confirms our first hypothesis (H1) that HGPE is a space for the debate of propositions despite elements of personalization and image building, which stand out in the literature on campaign modernization (MAZZOLENI, 2010), however this is a feature that diverges rather than converges by party. The PT's election advertising was consistently more thematic than the PSDB's, except in 2002. The professionalization of campaigning and political marketing has not been able to foster a pasteurization of advertising, to use the term coined by Miguel (2004), according to which that there is not much convergence between parties, which are important for defining the construction of political campaign advertising from the content point of view, even where there are many changes of scenario and candidacy. Graph 03. Discussion of themes of social welfare and economy by parties (%) Source: Elaborated by the authors, with data from the CPOP-UFPR.
Graph 04. Discussion of themes of social welfare and economy by parties (%) Source: Elaborated by the authors, with data from the CPOP-UFPR.
Specifically in respect the presidential contests played out by the PT and the PSDB in the last two decades -at least as far as party centrality and coalitions are concerned -and with reference to economic and social inclusion policies, as per the literature cited above, the data showed that there was no predominance of these themes in general for either party. The economy stood out in 2002 in PT campaign advertising when Lula argued that he would guarantee the maintenance of economic stability. From then on, PT campaigns showed the results of its social development policies while in government, with less space given to the theme of the economy compared to 2002. In the case of the PSDB, the economy received even less centrality than in PT advertising. Other economic matters can arise and be debated in HGBE as a matter of course, such as employment and taxes. Few of either party's HGPE segments between 2002 to 2014 were entirely devoted to macro-economics.

Discursive strategies: the importance of party positioning
The second characteristic of the HGPE that we analyzed concerns the strategies adopted by the parties 9 . Table 04 indicates that the PSDB made use of more 'call for change' strategies, with 19 percentage points more than the PT, and 'attack', with 21 percentage points more. These strategies reflect the positioning of the party in the campaigns and, therefore, the results were to be expected considering that the PT was the incumbent party in every election except the 2002 one. This is related to the proposition of Figueiredo et al. (1997), on how positioning affects strategy choices. In 2018, the parties assumed different positions, with the PT opposed to the incumbent president, Michel Temer of the MDB, and the PSDB having been part of the Temer Government since 2016, following the impeachment of Dilma Rousseff. According to Figueiredo et al. (1997), these are typical characteristics of challengers and, therefore, are more typical of the PSDB in the period that the PT. By 2018, the PT was no longer in power, given the 2016 impeachment and removal from office of president Rousseff.  From a longitudinal perspective, as shown in Table 05, it can be seen that PT built its strategies in different ways according to the election. This suggests that strategic differences are not partisan but depend on the position occupied by each party and the type of candidacy, as discussed below. In 2002, the PT employed a 'call for change' strategy, exceeding the average by more than 21 percentage points, because it was a challenger to the PSDB that had been in government for eight years.
In 2006, the PT's first election as an incumbent, the party campaigned using the 'use of office' strategy 36.6 percentage points more than the average. The party had occupied the presidency for four years and was therefore able to highlight the president's experience. In 2010, it used an endorsement strategy as its presidential candidate,  Source: Elaborated by the authors, with data from the CPOP-UFPR.
As Graphs 05 and 06 show, the two parties' strategies differ. This is because, unlike format or themes, strategy in an electoral contest is more related to the position and posture of the parties and candidates in each election than to any strategic tendency that may have gained prominence over time due to the modernization of political communication (MAZZOLENI, 2010). Even 'attack', which Borba (2015) shows to have grown compared to other strategies, stands out only when the need for its use is perceived from the context of the candidates and their position. Longitudinal growth can be seen by observing the PSDB and its significant use of 'attack' and 'call for change' strategies in every election. However, the same cannot be said of the PT, in which the most significant attack data only appear when the party is the challenger (i.e. in 2002 and 2018). We cannot therefore say that use of this feature has gained more visibility over time. Strategies vary between parties and over time without an apparent pattern. This is explained by the position of each party (incumbent or challenger) and the context of the election, since the longitudinal graphs indicate many changes. The scenario of the elections in each year and the relative positions of the candidates (challengers or representatives) can help us understand significant differences in use of HGPE rather than partisan or temporal considerations. Figueiredo et al. (1997)  Year-on-year variation can be seen longitudinally according to the position of each party and the characteristics of the candidates themselves. This does not indicate that one strategy has become more favored or that HGPE content production has changed over time. Context, candidate type and candidate's relative position are the variables that explain the differences that appear over time, and there are no linear changes between candidates or shared by them as they occupied distinct positions. Our second hypothesis (H2), that strategies vary according to a given candidate's position in an election (incumbent or challenger), is therefore confirmed.
In addition, non-linear variation occurs for all strategies and not just for 'attack' and use of political support.
Despite the importance of each party's context and relative position, one item demonstrates some stability among electoral advertising strategies employed by both parties: 'use of office'. Although more favored by the PT, 'use of office' was also central to PSDB strategy, even though it is typically used by incumbents (Figueiredo et al., 1999) and the PSDB was the challenger in three of the five elections. This is because the presidential candidate normally has a political career to demonstrate experience and credibility in administration. in dispute and we expect more debate and less confrontation than in the second round, when the election gets tough.

Final considerations
The purpose of this article is to investigate how the PSDB and the PT used their the HGPE in the five elections held between 2002 and 2018. It presents a comparison between the parties regarding the use of two characteristics -content and strategyand how time influenced the way these elements were used. Its main purpose was to identify similarities and differences between the parties, as well as to verify whether over time some characteristics became more or less absent from advertising, regardless of each party's choices, and to identify changes in the dynamics of HGPE use over time, as it has been an important element of political campaigning since at least 1989. The 1994 and 1998 elections were not considered for methodological reasons, since the database we analyzed only contains electoral advertising from 2002 onwards.
Of the five elections we looked at, the only once where the PT and the PSDB were not the main contenders was that of 2018.
The choices we made to create our research dataset did not allow us to address HGPE use in its entirety. The elections we looked at were first -past-the-post contests featuring large, professionalized parties, with substantial resources and ample HGPE time. Their shared characteristics could bring the PT and the PSDB closer to one another. Limited HGPE time also imposes different uses. Other analyses could certainly be made of other contexts not considered in the scope of this article.
Regarding both party and temporal differences, it is important to emphasize that we analyzed two characteristics only; i.e. we did not analyze format, appeal or language, all of which may also reveal divergences and developments in their use of advertising. Our objective was to focus on content. We limited our study to the first rounds of the elections we analyzed. This was necessary for our comparison since in the second round, when the parties have longer and equal HGPE, they can adopt strategies, formats and content that differ from those used in the first round.
Despite the necessarily limited scope of our research, our results provide important elements for the understanding of HGPE as a campaign tool. The extent to which HGPE is thematized depends on the relative position of the parties, as per our first hypothesis hypothesis (H1). Temporal change that might have transformed 'different' into 'equal' does not seem as relevant as party distinctions that are maintained over time. By contrast, the longitudinal distribution of themes does not indicate an increase or decrease for both parties, but only for the PT.
Therefore, the PT, in addition to being more thematic than the PSDB in each election, with the exception of 2002, shows significant growth in thematic HGPE. The PSDB oscillates between a focus on image and one on themes. Therefore, themes are a stronger campaign feature for the PT, one which proportionally gains more space year by year. In this sense, the differences between the parties are more explanatory than the temporal aspect for understanding the thematization of election campaigns. This is a finding that highlights the relevance of parties in Brazilian electoral contests. From 2002 to 2016, the PT made considerable investments in public policies to reduce social inequality. This was a topic that did not gain prominence in PSDB campaigns in the FHC era, but which has become central to campaigning since then. The economy tends to be highlighted in election campaigns that take place during recessions, with the two Our second hypothesis (H2) that strategies vary according to a given candidate's position in an election, with no significant temporal or partisan influence, was also vindicated by our findings, which showed that regardless of party and time between disputes, some strategies were repeated according to the positioning of the candidate. Thus, the variation of discursive strategy is seen election by election and occurs according to the party's relative position and the characteristics of the candidates themselves. This suggests that strategies do not 'grow' longitudinally -over time -and do not reflect change due to party matters.
Changes in discourse, it seems, occur over time and have no apparent direction or linearity. Therefore, they are not cyclical, as per Stimson's proposal (1976). Context, candidate type and candidate's relative position are the variables that explain the differences that appear over time, and there are no linear changes between candidates or shared by them as they occupied distinct positions. Disorderly variation within parties and over time occurs for all strategies and not only for some of them.
We can therefore say that the proposal of Figueiredo et al. (1997) remains relevant, at least for understanding and explaining the use of HGPE in national elections.
Thus, despite its limitations, we believe that this work makes two central contributions to the field. The first of these is a comparative contribution that explores the use made of HGPE by the PSDB and the PT. This is in line with other studies that seek to perceive party differences (CERVI, 2016;D'ARAUJO, 2011); here the focus has been on how they express themselves in election campaigns using HGPE. We also sought to make observations that went beyond just one election and allowed us to look empirically at the broader campaign-building process between the two parties over five elections. This longitudinal study constitutes research on political advertising and HGPE, which remains rarely mentioned in the literature.

Translated by Eoin Portela Submitted on October 14, 2019
Accepted on August 18, 2020