To understand the social political forces driving Brazil’s current election, Alto Data Analytics conducted an advanced analysis of the digital public sphere from August 20th to September 17th, 2018 in Portuguese. Two primary data sets were analyzed:

  • Data set one was comprised of public data sources: Twitter 75.9%, Facebook 17.5%, YouTube 5.8%, as well as blogs, forums and news media, from 3.7M authors generating a total of 34.1M results and representing up to 5,944M Potential Max Impressions.
  • Data set two was a sample of 50,180 paid Facebook digital display ads connected to the electoral campaign from Facebook’s advertising archives. This data set represented between 473M and 1.17M ad impressions, equivalent to an investment between R$4.2M to R$19.6M, a portion of the total campaign investment estimated to be a maximum of R$ 67.7 million during the period of analysis.

The new normal in digital electoral campaigns

Alto’s analysis of Brazil shows that the current electoral campaigns fit into the new normal that is being defined by six key trends in recent global elections:

  1. High polarization within debate.
  2. Candidate’s personality and leadership style capture more attention vs. political structures and parties.
  3. The importance of the female vote and the emergence of gender politics as campaign themes.
  4. Abnormal high-activity users, (i.e. beyond normal human behaviour when posting, replying, re-tweeting,) suggesting the use of automated or quasi-automated bots or strategies combining human and automation technologies.
  5. Users geo-located from multiple locations also influencing the debate.
  6. Big changes in the media landscape, with emerging digital news media platforms and non-Brazilian websites reaching significant levels of virality.

Highly polarized debate

For almost every 5 mentions of candidates, there is only 1 mention of a political party. Our analysis indicates candidates captured 26,313,837 results – 77.2% of total public conversation – compared to 5,428,296 mentions of any other political parties involved.

Personality vs. Political Parties

By analyzing the top 160 highest frequency terms interconnected from the public discussion through 2.67M messages, Alto’s analysts observed two distinctive groups of narratives: those supporting and those against Bolsonaro. Bolsonaro or themes related directly to him captured 51.2% of the highest frequency terms, while the remaining 48.8% connected to all other candidates.

Narrative analysis - top high frequency terms

Gender politics

Women were at the core of the debate and captured a significant number of the most frequently mentioned terms – this occured both in narratives pro and against Bolsonaro and dedicated terms and the top hashtags driving the campaigns:


Salient gender political narratives and their relevance in the public debate has been a consistent trend within the Facebook digital sample of 50,180 ads. Our analysis showed females from 35 to 65 and above, and those between 13 and 17 years old were subjected to higher intensity of Facebook ads impressions than males in the same age segments.

Facebook Ad impressions by gender

Abnormal high-activity users

To understand how public digital information propagates, Alto analyzed a Twitter network of 10,603,473 messages from 238,066 users with at least one of their messages re-distributed or re-tweeted. This network represented the core of the public debate as Twitter was the most active public data source. Alto’s data science team, then detected four key communities:

  1. Left Community: PT (Lula Da Silva & Fernando Haddad) and PSOL (Guilherme Boulos).
  2. Center-Left Community: PDT (Ciro Gomes) and REDE (Marina Silva).
  3. Center-Right Community: Partido Novo (João Amoêdo), MDV (Michel Temer) and PSDB (Geraldo Alckim).
  4. Right Community: Bolsonaro.

Center-left was the largest community: 138,802 users generating 1,780,305 messages, 13 per user. The Right community was most active in message output: with only 50,355 users generating 5,921,834 messages, 117 per user, which can be considered an early signal of abnormal level of activity.

Community Analysis: 4 key emerging communities

Bolsonaro’s digital impact was far superior to any other candidate on Twitter. Bolsonaro’s profile was the most influential as calculated using Alto’s proprietary algorithms for network topology analysis. His community gathered 6 of the top 15 most influential profiles within the network. PSL was the most influential political party in the digital debate too, but still below the level of influence reached by Bolsonaro.

Top 15 influencers among candidates

Alto Data Analytics detected abnormal high-activity Twitter users, i.e. beyond normal human behaviour automation. By analyzing users with or more than 4,000 re-tweets and users with activity between 1,500 and 3,999 re-tweets, data scientists estimated 30% of all re-tweets in the Right and Left communities were published by 1.3% of users. Bolsonaro’s Right community includes 663 users with this abnormal activity, while the Left community gathered 290 users with similar levels of high activity.

High activity users tend to group in Right and Left communities

Geo-locations of users with highest engagement

Focusing on those users with the highest engagement, analysts confirmed the majority of these users were geo-located in Brazil (55.5%) and that around 1/3 of them (37.9%) did not indicate geo-locations in their profiles. Other geo-locations of users included The United States (4,321 users), Argentina (1,739 users), Portugual (1,412 users), Spain (1,352 users) and Venezuela (785 users).

When analyzing users’ interactions, analysts discovered Argentinian and Venezuelan users tended to cluster and interacted more actively with the Left community. A small group of user geo-located in Portugal tended to be highly connected to Bolsonaro’s community.

Big changes in the media landscape

Analzying the role of media, data scientists identified the top 150 most viral domains, meaning the domains with the higest numbers of sharing. Alto found 331,626 users propagated these top 150 domains in 2,882,034 comments or messages. A deeper analysis surfaced these insights:

News and Political Party websites played key roles in the debate: 58% of top 150 domains belong to sites focused on News and 11.3% to Political Parties. Crowd-funding sites, podcasting sites and government sites were among the top 150 sites.

The top 30 most shared domains* indicated the number of times the domain was shared during our analysis period. Top domains include:

Domain Times Shared 362,816 288,606 266,996 157,615 145,609 134,463 123,773 108,541 71,522 52,563 50,675 49,939 37,272 34,892 33,738 31,363 28,740 27,756 26,133 25,857 24,156 23,718 22,243 21,404 18,935 18,162 18,102 18,030 16,842 16,634


Sites distributed by users belonging to Left or Center-Left communities include:,,,, and Right community users focused distribution on,,, and

Bolsonaro’s community is the only community with a non-Brazilian domain within its top-shared domains: Only on Twitter 8.430 users shared at least one piece of content from during the period of analysis. 

What is Gab defines itself as, “A social network that champions free speech, individual liberty and the free flow of information online. All are welcome.” According to Wikipedia “Gab is a Philadelphia, Pennsylvania-based social networking service formerly based in Austin, Texas. It was created as an alternative to Twitter and promotes itself as supporting free speech” “Gab has been described as a platform for white supremacists and the alt-right.” Within Alto’s analysis identified six narratives emerging in strong connection with Bolsonaro:

  1. #ImprensaCretina. Narrative presenting Bolsonaro followers as victims of traditional media that these followers label as the “fascist left”.
  2. “#PesquisasMENTEM”. Narrative supporting the idea that opinion polls are lying and cannot be trusted.
  3. Bolsonaro’s attack: news surrounding the assult and support from followers.
  4. Financial Agenda. Economist Paulo Guedes warns media that they will have to find new ways to finance themselves in case Bolsonaro wins  (#PauloGuedesNaGloboNews).
  5. Gab as a free space away from those willing to silence Bolsonaro and his supporters.
  6. Miscellaneous issues or references in connection with Bolsonaro: God, Family, Security, Less Brasilia and more Health and Education, Ending of PT.

When analyzing the users sharing or mentioning Gab, Alto’s data scientists found  95.9% belonged to Bolsonaro’s community and also mainly shared content from two domains: and

Part of the big changes observed in the media landscape are also connected with new  emerging media that have reached top positions in our domain analysis as Renovamí Renovamídia is a news site founded in March 2017 and run by one journalist, his wife and a group of supporters, with a clear vision, mission and values as described in its crowdfunding profile:

Pepe the Frog

During the analysis, Alto’s data scientists observed how groups of authors across multiple sources extensively used the icon of “Pepe the Frog”. Pepe the Frog is described by Wikipedia “as a popular Internet meme. A green anthropomorphic frog with a humanoid body, Pepe originated in a comic by Matt Furie called Boy’s Club. It became an Internet meme when its popularity steadily grew across Myspace, Gaia Online and 4chan in 2008. […] By 2016, the character’s image had been appropriated as a symbol of the controversial alt-right movement.

Users using Pepe the Frog enabled Alto to understand to which point there was a cultural transfer of this symbol to Brazil: on Twitter alone analysts found 3,335 authors producing 162,229 messages including Pepe the Frog, users also included it as part of their names or biographies. Authors with high activity included: with 9,061 messages and with 4,945 messages.

80% of users using Pepe the Frog also shared content connected with Bolsonaro’s community  including sites such as,,, and

The activity of these users steadily increased during the period of analysis from normal aggregated activity of 30,000 messages per day, to more than 60,000 messages per day by the end of this analysis. Their combined activity was continuous with almost no hour of the day left without posting.

When geo-locating them 91.9% of them were not geo-located and only 8.1% were geo-located in Brazil. This appeared to be a relevant deviation from the average in which 37.9% of  users did not indicate any geo-location in their profiles.

Interested in our work? For questions or a presentation of the complete analysis, please contact us at To learn more about Alto Data Analytics and our proprietary software, subscribe to our newsletter below.

*Methodological notes

Influence: Alto measured influence mathematically based on the structural connections of the identity in the digital debate and its ability to propagate its messages to larger audiences in the network (centrality measurement).

Domain analysis: Alto based the domain analysis in this research on Twitter, hence has been removed from the list of top domains. For consistency 4 other domains have also been removed to understand the visibility and propagation of media:,, and The direction and insights of the results do not change when including them in the analysis. However, understanding where those domains are within the top could provide better perspective of overall results: (position #1, 8,917,072 times shared), (position #2, 425,626 times shared), (position #5, 269,384 times shared), (position #13, 55,786 times shared), (position #24, 27,625 times shared).