What do State Institutions Say? Twitter as a Public Communication Tool During the Impeachment of Dilma Rousseff*

The objective of this article is to analyze the content of public communications in situations of political crisis. We aim to establish whether and to what extent there was partisan instrumentalization of the digital social media channels administered by State institutions during the impeachment of Dilma Rousseff. We analyzed all posts published between 2015 and 2016 on the official Twitter profiles of Brazil’s Chamber of Deputies, Federal Senate, Presidential Palace and Federal Supreme Court. We considered all posts with the words ‘impeachment’, ‘impedimento’, (‘impeachment’) ‘afastamento’, (‘impeachment’) and ‘golpe’ (‘coup’) (n=795). Our methodology comprised a combination of quantitative (descriptive statistics) and qualitative (content analysis) strategies. We found that the Federal Senate’s twitter profile most frequently posted the word ‘impeachment’, while the term ‘golpe’ was most frequently posted by the Presidential Palace profile. Over half of the publications fit into the category of ‘dissemination of news’. The Presidential Palace’s Twitter profile exhibited a distinct pattern of behavior, predominantly posting tweets associated with the ‘promotion of ideas and expression of positions’. Therefore, the Presidential Palace favored a communications strategy with a partisan bias. This research is relevant as it uses empirical data to discuss phenomena tied to studies on public and political communication.

, or do they seek to fulfill an instrumental role by aiming to favor the perspectives of leaders, governments and parties (BARROS and MIOLA, 2015;BUCCI, 2015;WEBER, 2007) We took a quantitative approach to measuring the frequency of the terms 'impeachment', 'impedimento' (impeachment), 'afastamento' (impeachment) and 'golpe' (coup) in posts on the profiles we examined. A quantitative stratagem of this type allows for a qualitative approach -made possible through Content Analysis.
Examining content allows us to diagnose the ways in which these entities handled the impeachment process in greater detail.
The article is divided into four sections. In the first one, we discuss the concept of State entities using social networking sites for public communication purposes. In the second, we detail the methodological strategies we used for this article. In the third section, we present our quantitative and qualitative data. In the conclusion, we discuss our findings and compare them with the relevant literature.

The use of digital social media by State entities for public communication
Public communication involves at least two dimensions: firstly, there is the normative dimension, according to which the primary purpose of State communication is to encourage civic participation and thereby improve service provision and social control through transparency mechanisms; the second dimension privileges strategic communication that seeks to construct public images with the goal of conquering or maintaining power niches (BUCCI, 2015;MIOLA and MARQUES, 2017 MARQUES, AQUINO and MIOLA, 2014;OTT, 2017;PARMELEE and BICHARD, 2011;ROSSINI and LEAL, 2012). Twitter has thus become an essential environment for announcing government decisions and, of course, for promoting public images (ENLI and SKOGERBØ, 2013;JACKSON and LILLEKER, 2011).
A considerable part of the studies examining Twitter as a political tool focus on the electoral communication strategies employed by candidates (GOMES et al., 2009;VERGEER, HERMANS and SAMS, 2013;WILLIAMS and GULATI, 2013 ENLI and SKOGERBØ, 2013;JACKSON and LILLEKER, 2011;JUNGHERR, 2016;LARSSON and KALSNES, 2014). There have been many research articles published that attempt to define the discursive differences characteristic of each communication strategy (BOR, 2013;LASSEN and BROWN, 2011;UTZ, 2009).

Methodological strategies
In order to answer the research questions presented in the introduction, a quantitative and qualitative study was proposed with the aim of verifying: 01. the frequency of posts dealing with one of the following keywords: 'impeachment', 'impedimento', 'afastamento' and 'golpe'; and 02. the content linked to these tweets.
Our hypothesis, derived from the literature, is that: 'The official Twitter profiles We started out by creating a general overview of each profile that considered its creation date and number of followers. We then manually collected posts from 2015 and 2016 that contained the words 'impeachment', 'impedimento', Tweet form (direct interaction with users via re-tweet, response, or direct mention of "@" by users) Dummy variable that identifies if a tweet interacts directly with other accounts. These interactions are recorded when an institution uses the '@' symbol to refer to other users or when it re-tweets or replies to a post. This variable relates to a specific characteristic of Twitter, i.e., the possibility of using interaction as a communication strategy. This includes using interaction with other users as part of measures aimed at segmenting the target audience or valorizing the views of supporters (HEMSLEY et al., 2018).

Tweet content
Variable divided into seven categories based on the main subject in each tweet. The categories set out below are mutually exclusive; we have included example posts for each category. Tweets that merely report on the general steps of the impeachment process, such as dates when certain formal steps in the process will occur. #impeachment -De 07 a 12 de julho -alegações finais da acusação. [#impeachment -July 07-12 -prosecution's closing arguments.] (Senate -06/23/16: https://goo.gl/sZ1LgD).

Media agenda
Invitation to follow radio broadcasts (statements, interviews) or to consult materials related to the impeachment trial sessions; announcements of when institutional representatives will give interviews or make statements. process, we achieved a Krippendorff alpha score of 0.795 -considered reliable by literature in the area (KRIPPENDORFF, 2004). It is important to point out that we only took into consideration the text published in each tweet, i.e., we did not consider any images or the content of any links.
In order to provide answers to our initial questions, the following topic includes, in addition to the descriptive frequencies, the chi-squared and Cramér's V statistical tests. The chi-squared test verifies if there is a statistically significant association between categorical variables; the Cramér's V test indicates the degree of this association. We also present the standardized residuals of the contents of the posts (allowing for the identification of differences between observed and expected values in a given distribution). Finally, we present the correspondence analysis -an exploratory data technique that allows for graphic visualization of categorical variables (HAIR JR. et al., 2009).      At times, the Presidential Palace profile also highlighted the support of other political personalities for Rousseff, tagging their profiles: "@jaqueswagner:

Presentation of data
(...) "Tenho convicção de que ele acha, como nós, que não há lastro para impeachment" ["I am convinced that he thinks, like us, that there is no basis for impeachment."] 5 . And "Usar impeachment para resolver problema político é afronta à democracia, avalia ministro @edinhosilva" [Using impeachment to solve political problems is an affront to democracy, in the opinion of minister @edinhosilva] 6 .       ______________________________________________________________________________________________ 10 The standardized residual permits a direct comparison between values; it identifies the difference between observed and expected values in a distribution. In the confidence interval we applied of 95%, any value beyond +/-1.96 is statistically significant (FIELD, 2005).  In the Chamber of Deputies' profile, the type of content most posted was 'dissemination of news', with 78 tweets (about 59% of this profile's total). However, the most powerful theme was 'media agenda', with a standardized residual of 2.8; posts containing 'positioning' were the weakest (s.r. -3.2). It is therefore possible to infer that the Chamber of Deputies attempted to avoid passing value judgments about the process.
As stated earlier, the Federal Senate profile published 477 tweets mentioning at least one of the four terms in the reporting period, representing about 60% of the total corpus. In the case of the Senate, the most prevalent content was that related to 'dissemination of news', with 240 tweets (about 50% of the total   Although the other three profiles adopted a different Twitter utilization strategy (in terms of the frequency of publications, interactions with other users and the nature of posted content), we noted that the Chamber of Deputies and the Federal Supreme Court emphasized the dissemination of news. The Senate account tended to post on the steps in the impeachment process. That is to say that the predominant categories of content published by these three accounts are related to normative public communication.
The profile of the Presidential Palace, especially in the period before April 2016 (Rousseff was temporarily removed from office in May of that year), adopted a defensive position vis-à-vis the impeachment, with most of its tweets falling into the category 'promotion of ideas and expression of positions'. The Presidential Palace was the only of the four entities that presented more content related to 'positioning' than 'news'. This demonstrates how the Presidential Palace was more concerned with criticizing or coming out against the situation, especially by using the term 'golpe'. Moreover, the fact that the Presidential Palace profile stopped mentioning the impeachment as soon as Rousseff was suspended reinforces the impression that the profile had been politically instrumentalized. The Presidential Palace profile's behavior, therefore, reinforces the diagnosis of instrumentalization of public communication.
It is not surprising that the Presidential Palace profile adopted the stance identified here -especially because it is more directly linked to the construction of the image of the then president and her administration. profile. We therefore signal the need to investigate the possibility that political administrators pay more attention to certain communication platforms than others.
None of the tweets published by the STF was concerned with promoting interactions. The Chamber of Deputies followed the same pattern and did not mention or replicate content from other accounts. The Senate interacted with other profiles with some regularity, but the focus was on giving visibility to the words of several senators -particular attention was given to the profile of Senator Anastasia, the lower house impeachment rapporteur. The Presidential Palace also dedicated part of its posts to interaction but in most cases associated such interaction with a political position -giving visibility to Dilma Rousseff's personal profile.
In We cannot ignore the fact that the communication structures available to each State entity may vary, directly affecting how they respond to crises. In this sense, the performance of State entities depends not only on external pressures faced by them and their leaders but is also -as Greening and Gray (1994) argueassociated with the amount of resources (personnel and financial) allocated to them in a given situation.
Before concluding the article, we would like to point out that this work's novelty lies in its study of State institutions' use of social media, thus contributing to the literature on public communication via digital social media. The analytical framework designed and presented here may be useful for other works that seek to examine moments of severe institutional crisis. The data we obtained are consistent with the separation of public communication into two dimensions, as practiced by previous authors (MIOLA and MARQUES, 2017;WEBER, 2007). Moreover, it is necessary to study the use of these communication channels at times other than during political crises; analysis of these at other times may lead to different findings from those presented in this study.

Translated by Fraser Robinson
Submitted on October 20, 2018 Accepted on June 08, 2019