All the President’s Trolls: Real and Fake Twitter Fights in El Salvador
All the President’s Trolls: Real and Fake Twitter Fights in El Salvador
El Salvador President Nayib Bukele speaks at a news conference during a nationwide quarantine as the government undertakes steadily stricter measures to prevent the spread of the coronavirus disease, in Ilopango, El Salvador May 18, 2020. REUTERS/Jose Cabezas
Commentary / Latin America & Caribbean 12 minutes

All the President’s Trolls: Real and Fake Twitter Fights in El Salvador

The plunging homicide rate in El Salvador has sparked debate about the role of the new president’s hardline policies. Much of it transpires on Twitter, where his champions and critics engage in rows that could pre-empt reasoned discussion of how to keep tamping down violence.

El Salvador’s Nayib Bukele won the presidency in 2019 by promising to reduce the country’s then sky-high murder rate and tackle corruption. Homicide rates have indeed fallen sharply since his election. But Bukele’s policies have proved controversial. Critics say the president’s actions, such as cramming gang members into cells without daylight, and strong-arming the parliament and high courts, violate human rights and erode democracy. At the same time, these policies have made him more popular than ever, with many Salvadorans attributing the drop in homicides to his no-nonsense approach.

The conflict between these two camps is fierce. One of its main battlegrounds is online, particularly on Twitter. Bukele relies on tech to address citizens directly, and to that end he regularly takes selfies and posts memes (at one point sharing a picture of himself in a spaceship). Even though fewer than half of Salvadorans have regular access to the internet, Twitter has become so central to Bukele’s public messaging and government operations that The Economist recently ran a story about him titled “My Tweet is Your Command.”

Focusing on competing hashtags, both sides of the political divide engage in ferocious online slanging matches.

Crisis Group has documented a concerted effort from both Bukele’s supporters and his opponents to shape the online narrative around his more controversial policies, in part through artificial means. Focusing on competing hashtags – #BukeleDictador (#DictatorBukele) and #QueBonitaDictadura (#WhatALovelyDictatorship) – both sides of the political divide engage in ferocious online slanging matches. The result is to present Salvadorans with artificially polarised choices: reject Bukele, despite his apparent successes; or support him, and ignore the abuses committed by his government. The reality, as we show in our new report Miracle or Mirage? Gangs and Plunging Violence in El Salvador, is more complicated and nuanced. But as social media-fuelled polarisation intensifies, the risk is that both sides will shun the complexities of tackling gang violence in an effort to win the online popularity contest and the forthcoming elections in February 2021.

The Hashtag Wars

Part of Bukele’s approach to El Salvador’s security and health challenges is undoubtedly hardline. Shortly after taking office, he deployed joint military-police units to fight gangs and tightened restrictions on prisons, including forbidding family visits. He ordered the armed forces to occupy the legislature in a failed attempt to coerce it into ratifying a loan to fund his security strategy. After a temporary rise in homicides, Bukele backed using “lethal force” against gangs. The administration posted images on Twitter of near-naked gang members chained to one another. COVID-19 presented a new threat. Bukele has responded by detaining citizens who disobey the strict nationwide curfew in crowded “containment centers”.

International and domestic critics have responded with a stream of criticism calling Bukele a “dictator”. Still, many Salvadorans view these uncompromising policies as responsible for the reduction in homicides and the country’s relative success at staving off COVID-19. As a result, Bukele still has an approval rating of nearly 90 per cent. Our new report suggests that reality is more complicated. In fact, the declining murder rates may owe not only to the tough measures Bukele publicises but also to changes in the gangs themselves and fragile non-aggression pacts between them and government officials. These nuances are missing entirely from the online fracas.

A concern is whether these hashtags are artificially boosted

Two competing hashtags used on Twitter shine a light on these battles. #BukeleDictador first trended after the occupation of the legislature in February, and again in response to the president’s handling of COVID-19. #QueBonitaDictadura was deployed to fight back against this hashtag and negative press more broadly. To analyse the hashtags, we looked at posts (original tweets, replies, and retweets) through Twitter’s standard API, a set of procedures allowing access to the platform’s searchable data. Since this method restricts us to a week’s worth of posts, for #BukeleDictador we were able to collect 29,948 tweets posted between 27 April and 9 May 2020. For #QueBonitaDictadura, we pulled out 33,251 posts since its first use on 28 April until 9 May.

Documenting Manipulation

A first concern is whether these hashtags are artificially boosted. Inauthentic activity on Twitter takes many forms, from bots (automated accounts) to “sock puppets”, human accounts with deceptive online identities. Such manipulation is often difficult to detect, as no one metric definitively proves an inauthentic account – some people just use Twitter oddly. Accusations of manipulation in El Salvador have largely focused on the use of troll or net centers, involving paid humans running accounts to spread certain messaging. Bukele himself was implicated in a troll center case targeting newspapers in El Salvador, and recently the government accused the left-wing opposition Farabundo Martí National Liberation Front (FMLN) of running a troll center.

There is evidence of suspicious behavior associated with both hashtags. Figure 1 shows posts per hour for each. For both hashtags about 75 per cent were retweets, higher than for most organic traffic (meaning regular user activity). Since retweeting is easier than crafting original posts, a high percentage of retweets is correlated with manipulation. The volume and timing of posts (both retweets and original posts) for the two hashtags otherwise look very different. #QueBonitaDictadura peaked within four hours of its first use, and dropped off almost completely within a week. Though we do not capture its peak, #BukeleDictador was tweeted one to five thousand times a day, a rhythm that continued even in late May.

Use of both hashtags points to manipulation in various ways beyond the high proportion of retweets. Some 4.4 per cent of #BukeleDictador posts and 5.6 per cent of #QueBonitaDictadura posts were from accounts deactivated by the end of May, a signal that Twitter may have found users suspicious. The Coefficient of Traffic Manipulation, which measures deviation from regular traffic, similarly points to suspicious activity. Using recent activity and user characteristics, a number of measures try to score accounts according to the likelihood that they are bots. Depending on the method and threshold we use, estimates for the percentage of tweets produced by bots range from very low (15 per cent) to quite high (70 per cent) – the enormous variance shows how challenging it is to identify non-human users. But whichever method is used, the estimates of how many tweets are generated by bots are always similar for the two hashtags. Both appear to profit more or less equally from fake, or non-human, tweets.

#BukeleDictador was propagated by a small number of accounts. Fully 62 per cent of tweets came from 500 users with suspiciously high rates of tweeting: in one week, they collectively posted more than 300,000 times, well in excess of standard estimates for suspicious activity. Two accounts posted more than 3,000 times in two days. Some 26 alleged opposition trolls have been outed by the administration. Eight are no longer on Twitter, but the visible histories of the remaining users in the list show that they dramatically increased posting in late 2019 (in some cases after long dormant periods). In violation of Twitter rules, these accounts frequently copy-pasted themselves – repeating the same exact message, but directed in reply to different users – and occasionally copy-pasted each other. At least two accounts use profile pictures of other people passed off as their own.

Bukele’s supporters, meanwhile, appear to have been building a network of pro-government accounts since the week he took office. Figure 2 shows the date of creation for accounts that tweeted the two hashtags: there is an unmatched spike in #QueBonitaDictadura users joining Twitter in the weeks directly after Bukele’s inauguration on 1 June 2019. This is suspicious. We would expect a smoother bump if supporters joined Twitter to champion Bukele, given that he had been preparing to take office for close to four months and this spike began several days after his inauguration. While other salient events – Bukele’s election, for example, or the military occupation of the legislature – also saw an increase in accounts created, these bumps are smoother and dwarfed by the inauguration. In one day (5 June), nearly as many pro-Bukele accounts were created as in the entire month leading up to the inauguration.

For both hashtags, many of the accounts posting them were created recently, which tends to signal inauthentic behavior: as accounts get deactivated, new ones crop up to replace them. Some 8 per cent of #BukeleDictador and 14 per cent of #QueBonitaDictadura accounts were created in just the previous two months. These users were responsible for 10 and 13 per cent of tweets, respectively.

Figure 2: Date Twitter Accounts Were Created Crisis Group analysis of Twitter data

Amplifying Elites

Though these hashtags were spread in part through artificial means, real political elites set the content. #BukeleDictador gained popularity after being posted on Twitter by deputy Alexandra Ramírez, whom the government alleges coordinated the FMLN troll center. The role of political leaders in designing opposition rhetoric was also visible in another case, when right-wing opposition politicians and a number of other accounts with large followings posted identical, copy-pasted tweets over the course of an hour.

#QueBonitaDictadura was first posted by Porfirio Chica, a communications and public relations strategist who ran a “secret network” and “propaganda machine” that aimed in 2015 to reelect ex-prosecutor Luis Martínez, who is now in prison on corruption charges. Chica claims that his services to Martínez were free, but the website El Faro discovered text messages from him on the former prosecutor’s phone referencing costs. Chica is tightly linked to Bukele; Foreign Policy identified him as a campaign consultant, and The El Salvador Times called him a member of the Bukele campaign’s “circle of trust”.

Just after 7pm on 28 April 2020, Chica tweeted three times in quick succession with the same structure: #QueBonitaDictadura alongside a picture of the security services helping an elderly or disabled Salvadoran. #QueBonitaDictadura was tweeted or retweeted 621 times over the next two hours, including once, shortly after 8pm, by Bukele. At 9pm Última Hora, a digital publication owned by Chica, posted a story that #QueBonitaDictadura reported wide support. Over the next two hours #QueBonitaDictadura was tweeted or retweeted 4,324 times, including eight retweets from the president in a five-minute window.

A version of these dynamics plays out among elites at large. Among our data we identify 54 opposition and 102 pro-government elites who are either verified or have more than 2,000 followers. Elites were responsible for less than 1.5 per cent of posts, but they were disproportionately influential: 19 per cent of #BukeleDictador posts and 31 per cent of #QueBonitaDictadura posts retweeted them. Elites played a particularly influential role in getting #QueBonitaDictadura trending, with most of their posts in the hours directly after its first use.

These hashtags propagated two polarised interpretations of Bukele’s policies: that they are either eroding basic liberties or bringing peace.

International users also play a role in propagating the hashtags, particularly #BukeleDictador. Relying on self-identified location, roughly equal proportions of both hashtags’ users claimed to be living in Europe, the U.S. or Canada. But 12.5 per cent of #BukeleDictador accounts self-identified as Cuban, Venezuelan or Nicaraguan. Bukele broke from the FMLN when he ran for president, and the vast majority of these accounts are dedicated to propaganda for current and former left-wing authoritarian leaders in Latin America, such as Fidel Castro, Hugo Chávez, Nicolás Maduro or Daniel Ortega. The most retweeted #BukeleDictador post among the sample came from an anonymous user claiming to be in Nicaragua, describing herself as “a Sandinista, revolutionary woman, Nicaraguan, chavista, communist, leftist”. This suggests #BukeleDictador extended to a network of international, and often suspicious, accounts.

Competing Narratives

These hashtags propagated two polarised interpretations of Bukele’s policies: that they are either eroding basic liberties or bringing peace. Some of the most retweeted #BukeleDictador tweets accuse the president of violating human rights and democratic values:

  • “Prohibited from public transit. Prohibited from thinking. Forbidden from dissenting. The master of El Salvador has again broken the law and declared a State of Siege. Forbidden from crossing municipal borders. He will decide when you can buy food. #BukeleDictador God will judge you one day, tyrant”. (Retweeted 454 times)
  • “People also loved Hitler and Pablo Escobar. They were popular, but this doesn’t mean one wasn’t a massive genocidal psychopath and the other a murderer and drug trafficker. So stop with the f***ing popularity. #BukeleDictador” (Retweeted 238 times)

#QueBonitaDictadura instead depicted Bukele as the purveyor of peace and security. Some19 per cent of #QueBonitaDictadura posts included one of ten photos of the military providing services, such as helping an old man carry bags of rice. Google Image Search shows that several of these images originally appeared in past tweets from Bukele’s communications department.

Activating Online Networks

#QueBonitaDictadura and #BukeleDictador are two online rallying cries in a broader conflict over the president’s policies and style of governance. Closer analysis of accounts that make use of one or the other of these hashtags shows that they are part of a consistent pattern of using Twitter to promote or denigrate the president. To explore this behaviour, we pulled together the history of the 500 accounts that tweeted #BukeleDictador most often, which we consider suspect anti-Bukele accounts, and the 542 #QueBonitaDictadura users who joined directly after Bukele’s inauguration, which we consider suspect pro-government users. So that we could explore behaviour over time, we included only accounts that tweeted fewer than 3,200 times in 2020, the maximum number of posts we can view through Twitter’s API.

Opposition accounts paint Bukele as corrupt (#WhoPaidForTheOsirisTrip, a reference to a payment scandal), ineffective (#IncompetentGovernment), and authoritarian (#WeWillDefendDemocracy). The hashtag #WhereIsBukele smeared the president for his perceived absence during the coronavirus crisis, which prompted the meme of him in a spaceship (Bukele himself posted, “the rumours of my abduction by aliens are totally unfounded” in response). Pro-Bukele accounts’ most-used hashtags promote the president’s policies (#TerritorialControlPlan, referencing the president’s security policy), smear the opposition (#ReturnWhatWasStolen, a rallying cry against corruption) and show support for him (#I’mWithBukele).

Not surprisingly, both sets of accounts appear most active when Bukele faces a significant crisis. Figure 3 shows variation in tweeting among the sample of opposition (left) and supporters (right). Daily posting is visually represented as a percentage of the mean of each group – since the two sets of users are very different, this allows a better comparison of how their tweeting patterns change over time. The accounts tweeted steadily until a spike in posting just before Bukele’s February occupation of the legislature. Supporters’ top hashtags during this period were #PlanControlTerritorial (#TerritorialControlPlan) and #ElPuebloManda (#ThePeopleRule); opponents’ were #LaDemocraciaSeDefiende (#WeWillDefendDemocracy) and #BukeleDictador. The onset of the global COVID-19 pandemic then led to a sustained period of higher engagement.

Figure 3: Tweeting History for Sample Anti- and Pro-Bukele Accounts Crisis Group analysis of Twitter data

Tweeting for El Salvador

Bukele’s supporters and opponents are both immersed in a sustained effort to shape the online narrative around the president’s policies, including through possible platform manipulation. Though less than half of Salvadorans have regular access to the internet, given the centrality of Twitter to Bukele’s government – he announces firings, establishes policies and criticises opponents on the platform – it is perhaps not surprising that the two camps have taken their fight online. But inauthentic activity may have worrying consequences for how both domestic and international audiences interpret politics in the country.

Our findings show that Bukele’s opponents and supporters use Twitter as a tool to inflame, polarise and simplify political debate in El Salvador.

Our findings show that Bukele’s opponents and supporters use Twitter as a tool to inflame, polarise and simplify political debate in El Salvador. Opposition concerns about Bukele violating democratic principles and civil liberties are important, but labelling him a “dictator” is an exaggeration that overlooks his accomplishments. Supporters present Bukele and his security service as forces dedicated to peace and the public interest, but whitewash their abuses of power. Starkly polarised takes on these life-and-death issues are hardly conducive to responsible decision-making and compromise, and instead seem to encourage legislative gridlock at a time when Bukele and his opposition should instead engineer long-term strategies to cement the welcome decline in homicides.

Elections in February 2021 will determine how many of Bukele’s supporters enter congress, and – as in the rest of the world – the battle for votes is happening partly on Twitter. The fact that at least part of this online battle is the result of artificial amplification of messaging on both sides adds to concern that the Twitter debate is stoking a climate of hostility that serves partisan political interests. This distracts from a very necessary debate about how to cement Bukele’s successes while moving away from “iron fist” policing and toward policies aiming at preventing gang violence.

Miracle or Mirage? Gangs and Plunging Violence in El Salvador

Subscribe to Crisis Group’s Email Updates

Receive the best source of conflict analysis right in your inbox.