Brazil 2018: Who’s winning at internetting: Bolsonaro, Haddad, or Ciro?

Millions of Brazilians are taking to Google and social media apps to find out more information about their presidential candidates as we approach the first round of a presidential election in Brazil. For a primer on how presidential elections in Brazil work, take a moment to read this and learn more about the leading candidates here. Yes, it’s complicated, but the context will help with everything that follows. For the purposes of this discussion, I have included data about candidates Ciro Gomes, Fernando Haddad (with data about Lula as a supplemental proxy), and Jair Bolsonaro.

There are several well-known surveys circulating in Brazil that show voter intent. While these are useful studies that provide a baseline for understanding the likelihood of a certain election outcome, the information that an individual finds on the first page of Google results has a profound effect on consumer and voter choice.

This study shows that search engine result ranking is proven to sway consumer opinions. For instance,  information found high up on the first page of Google results is more likely to be trusted, whereas lower-ranking content is less trusted.

Biased search rankings can subsequently shift the preferences of undecided voters by 20% or more. This will play a significant role in an election with a large body of undecided voters.

This election proves to be very unpredictable because of a few factors. Here are a few:

  • Leading candidates have exceptionally high rejections rates. How will that play out?
  • There is a tendency to lean into more polarized political positions, as has been observed in recent elections around the world, exemplified well by the 2018 US presidential election.
  • Since the most popular candidate, Lula (Workers’ Party aka PT) is barred from running for what appear to be trumped up corruption charges. He is now a political prisoner. His party is relying on a transfer of votes to their more recently proclaimed candidate, Fernando Haddad. In the context of the deep divisions of 2018, it is not known how this will impact things.
  • Disillusionment with the political system and politicians in Brazil is very high.
  • People may feel secretive (or guilty?) about supporting Jair Bolsonaro given the high rejection rate and might be keeping their vote a secret, therefore skewing polling results. This would give him a clear lead in actual results that are not reflected in polls.

Is Bolsonaro, in the words of Renata Gomes, the “imperfect emissary of the silent majority”? That’s one possibility that this data may confirm — that despite polls and surveys, his fast victory may be lurking in the shadows of so-far publically unspoken, majority support. If so, here are some thoughts on that.

While I do not want to be correct about this, it does look at the moment that Bolsonaro is winning at “internetting”, and I would predict he is positioned to win the Brazilian presidential elections in either the first round (although pretty unlikely?) or second round (high likelihood) because of the impact that search results and trends, plus social media activity, will have on voters when combined with the complicating factors mentioned above. Oh, and then there’s the fact that Steve Bannon has been voluntarily consulting for Bolsonaro, which has probably helped him nail down his rhetoric and digital strategy.

Let’s take a look at some metrics to understand why I (still very hesitantly) think that he can win. To reiterate, I have included data about candidates Ciro Gomes, Fernando Haddad (with data about Lula as a supplemental proxy), and Jair Bolsonaro. In this analysis, we are assuming a best-case scenario for Haddad, where the vast majority of Lula votes would transfer to him.

Ultimately, if no candidate gets more than 50% in the first round, we can expect a Haddad vs Bolsonaro contest in the second round (updated results here on Bloomberg) where the surveys show a tied race, promising a narrow victory in either direction (as of October 2, 2018).

Using search volume and search trends as a measure of Brand Awareness

Using these search trends as markers of brand awareness, the emergent trend during the past 30 days indicates an increasingly consistent interest in and curiosity about Bolsonaro. In fact, I would even say that he is leaving others in his shadow in recent weeks. But in this measure, it’s clear he’s dominating.

Search trend graphs illustrate the search advantage that each candidate has gained/lost as the defining events of this year’s race have unfolded. Most recently, you can see the advantage that Bolsonaro has.

Google Search Trends: Four candidates from the past 12 months
Google Search Trends: Zooming in on the past 30 days

The biggest spikes correlate with the following events:

April 7, 2018 – Lula turns himself in to federal authorities to begin serving a 12-year prison sentence for corruption.
August 15, 2018 – Lula submits official registration to be the Workers’ Party candidate in the upcoming presidential election.
September 1, 2018 – Brazil’s highest level electoral court bars Lula from running because of his conviction.
September 7, 2018 – Bolsonaro gets stabbed at a public campaign rally, causing him to get international attention and step back from public appearances and debates until he is healed enough to get back to work. During this time, he has maintained an online campaign.

Monthly Search Volume:

Another measure where Bolsonaro leads by a huge margin is the sheer volume of monthly searches. While the previous trendlines shows spikes and dips in the search tendencies, this metric (monthly search volume) is an indicator of both curiosity and increasing awareness. Even when pairing this with a high rejection rate, there is a huge opportunity here for search results to influence voter decisions, especially for the neutral, curious, or undecided voter.  Average search volume over the past month follows the trend of August. Not surprisingly, Bolsonaro leads in both cases.

12-month average
Bolsonaro: 1,000,000
Ciro: 301,000
Lula: (unavailable from Google Ads keyword tool at the time of writing)
Haddad: 33,000

August 2018
Bolsonaro: 6,423,370
Ciro: 1,599,940
Lula: 1,660,820
Haddad: 583,020

Search Result Sentiment and Auto-fill.

In the first page search results,  sentiment for all of the leading candidates leans neutral/negative. No candidate seems to get more positive coverage than the other in their ‘top articles’ news section respectively, and I would argue that this trend follows suit with a divided audience and a high rejection rate. Like the US political situation in 2016, constant, dramatic headlines are hammered out to generate clicks, views, and engagements in a competitive online news environment.

That said, a few things stand out in Bolsonaro’s favor that other candidates simply don’t have. The news of Bolsonaro’s stabbing seems to be helping to position him as a victim rather than a macho aggressor (which he is). Also, some of the highest ranking results in a query for Bolsonaro show that Brazil’s supreme court (STF) decided not to take the case against him for hate-speech, basically excusing him of responsibility for purveying racism, violence, homophobia and misogyny.

That is incredibly helpful for his campaign’s position and standing, given that if he were tried and convicted, he would be banned from running. It also further lends credibility to the narrative that he’s strategically developed of himself as a victim/underdog/outsider, despite being a politician for decades.

Another element of searching that has a tendency to reflect society and impact our thinking are google’s autocomplete results in the search bar. Here are the autofills in a neutral browser:

Acceptable Translations:

Lula imprisoned, Free Lula, Lula president, Lula candidate, Lula freed, Lula haddad, Lula 2018. (In addition to a few fun ones like “squid mollusk”, “colossal squid” and “giant squid”, since lula means squid, too!)

Haddad proposals, Haddad PT, President Haddad, Haddad charges, Haddad mayor, Haddad corruption, Haddad research, Haddad vice, Haddad investigated (Plus one for a company by the name, Haddad uniforms)

Bolsonaro recent news, Bolsonaro campaigns, Bolsonaro research/study, Bolsonaro 2018, Bolsonaro (political) party, Bolsonaro on Ratinho (show), Bolsonaro president, Bolsonaro died, Bolsonaro today, Bolsonaro on Globo news (tv network)

Ciro Gomes, Ciro Gomes corruption, Ciro Gomes aggression wife, Giro Gomes number (candidate ID), Ciro sales, Ciro Gomes site, Ciro Gomes machista, Ciro machista, Ciro Gomes campaigns, Ciro Gomes SPC (Credit Protection Service, referencing one of his proposals)

As you can see here in searches for Bolsonaro, despite leaning homophobic, a supporter/defender of Brazil’s military dictatorship, racist and a sexist, those words do not show up anywhere. In contrast, his autofills show lots of neutral curiosity and interest. His competition, however, is saturated with words that are more likely to negatively impact sentiment. This effect is both culturally reflective and persuasive, as explored here in this ad campaign by It will undoubtedly impact the outcome.

Reading Time

When it comes to “reading time” as yet another indication of openness/interest to candidates and their positions, Bolsonaro leads. According to a study by the popular content marketing ads platform, Taboola, (cited here in Portuguese, because it is not available in English) looks at how much content exists online and how many hours are spent consuming it by internet users.

Jair Bolsonaro – 41.19 million online articles and 1.14 million hours spent reading
Ciro Gomes – 25.72 million online articles and 777,000 hours spent reading
Geraldo Alckmin – 20.37 million online articles and 446,000 hours spent reading
Fernando Haddad – 17.65 million online articles and 594,000 hours spent reading
Marina Silva – 14.65 million online articles and 339,000 hours spent reading

Completing the digital puzzle: Social media audience and engagement alongside August Google search volume.

Here’s where I think the deal gets sealed. In 2018, social media engagement plays a huge role in every aspect of our lives (political or otherwise). Below we see the relative audience sizes, search volumes and engagement indexes from recent social media posts on candidate-owned accounts (samples taken in mid-September 2018). If I were to guess the outcome based on these numbers, the winner would be clear, again despite the obvious shortcomings, homophobia, misogyny, and frightening authoritarianism. The question remains regarding how voter intent, rejection, and opinion data intersects with this data to deliver a final result in real time.

For instance, how will the #EleNão (#NotHim) movement factor into the upcoming election results? Will Brazil really elect someone who has said and threatened so many horrible things? What does that really say about society? Do people now sympathize with him as a victim? Is this a bait-and-switch game waiting for moderates who might cast a vote for him in hopes he can “resolve crime”, but who might otherwise not be bigoted?

A final note: As this divide deepens and the national scene becomes more unpredictable, much more inspiring things can be found happening at the local level, in cities like Rio de Janeiro, where lots of candidates from the favelas are running for political office. Real change is ultimately coming, but it’s coming from the ground up, not top down.