Dubious voting, such as vote buying, prevalent on X/Twitter, contributed to an inaccurate portrayal of Trump’s electoral success.
Biased Social Surveys on X/Twitter
Informal political surveys carried out on X/Twitter during both the 2016 and 2020 U.S. presidential elections were significantly influenced by questionable votes, many of which might have been acquired from troll farms. This finding, as concluded by a group of researchers led by Przemyslaw (Przemek) Grabowicz, an associate professor at the University of Massachusetts Amherst, reveals that X/Twitter’s survey system intentionally presents biased public vote counts. On average, the outcomes of such questionable surveys favored Donald Trump over Joe Biden, 58% to 42% in a direct comparison, during 2020.
Evidence of Dodgy Survey Tactics
The team also discovered that there were about 50% more suspicious votes in pre-election surveys compared to those after the presidential elections, indicating that skewing social surveys is a deliberate strategy to impact political results. In 2024, such partial social polling from X/Twitter was overtly showcased by Trump on his social media platform, Truth Social, presumably to project his overwhelming popularity. These results were recently released in the Journal of Quantitative Description and will be published by the Association for the Advancement of Artificial Intelligence.
Scrutiny of Survey Authenticity and Influence
“During the 2020 presidential election, over 20 million votes were cast in more than 100,000 straw polls on X/Twitter,” states Grabowicz. “These polls collectively indicated Trump winning by a huge margin, when in reality, Biden clinched the election. We wanted to dig deeper to assess the legitimacy of these polls and what they could reveal about how social media impacts American politics.”
This was no small feat. “We estimate there are over 1 million social surveys of various kinds on X/Twitter every month,” mentions Grabowicz. These social surveys could cover any topic—preference for cats versus dogs, jazz versus heavy metal—and hence the researchers eventually filtered their outcomes to specifically political surveys posing some form of the query, “Who will you cast your vote for?” or “Who will secure victory in the election?,” and which listed both Biden and Trump (or Clinton and Trump for the 2016 election) as candidate choices.
Influence of Sex and Beliefs on Survey Outcomes
What they unearthed was remarkable.
Social surveys consistently anticipated a sweeping election triumph for Trump in both the 2016 and 2020 presidential races. On average, the 2020 social surveys had Trump leading by 58%, although he actualized only 46.8% in the presidential election.
Both the 2016 and 2020 social media surveys were primarily formulated by males who exhibited a clear preference for Donald Trump. In comparison to traditional exit polling conducted on election day, social surveys were more than twice as likely to be authored by males than respondents to exit polls. Furthermore, the political ideology of those creating and responding to social surveys skewed rightwards, while those who retweeted and liked social surveys were even more predisposed—by over 10 times—to identify as conservative.
But political alignment alone did not clarify what the research team observed. Interestingly, it seemed that Polish politics might have played a role.
Foreign Influence and Discrepancies in Survey Reporting
In 2020, the Polish state media TVP INFO published a comprehensive article on the outcomes of an X/Twitter survey it conducted asking participants who had emerged victorious in a Polish presidential debate. What TVP INFO alleged is that out of 35,202 votes, 19,539, or 44.5%, had been purchased from troll farms, which are easily identifiable with a quick online search.
“We aimed to investigate if a similar situation was unfolding in the U.S.,” mentions Grabowicz.
The main concern is a variance in how X/Twitter exhibits survey votes. There is a public figure—that anyone engaging with or voting in the survey can view—but there is also a private figure, accessible solely to the survey author. In the case of the TVP INFO’s Polish presidential debate survey, the public count was nearly 20,000 votes higher than the private figure—but the public had no means of knowing this.
In essence, the public cannot differentiate between a purchased vote and a legitimate one.
Revealing Concealed Discrepancies and Manipulative Approaches
To examine if a similar scenario was transpiring in the U.S., Grabowicz and his colleagues conducted their own survey asking participants whom they would vote for: “Potoo from Arizona, Walrus from Alaska, or Sheep from New York,” and then they bought votes for their own survey from one of the troll farms.
Upon analyzing all the data, the researchersdiscovered that the inconsistencies between public and private vote tallies were closely, yet not perfectly, aligned with the number of bought votes.
According to Grabowicz, “In some way, Twitter is eliminating the purchased votes from the poll creator’s perspective, but the method or reason remains unknown.” He promptly mentions that, without access to X/Twitter’s internal source code and data, he is unable to verify that all suspicious votes are indeed purchased votes.
Findings from Studies on Manipulation of Pre-Election Polls
The investigation also involved a survey of 984 creators of 2020 X/Twitter social polls, requesting access to their confidential vote count. Despite receiving responses from only a few, the outcomes were uniform across all the polls examined. Notably, there were approximately 50% more questionable votes prior to the 2020 presidential election compared to after, indicating a systematic effort to distort voter perception of public sentiment through social poll manipulation.
Opacity in Social Media Platforms
Moreover, some of the unreliable pre-election social polls forecasting a landslide victory for Trump were utilized to bolster beliefs in voter tampering once the official election results were announced. Trump himself recently posted a screenshot on Truth Social showing a biased ongoing poll on X suggesting that 70.1% of participants endorsed him.
Based on social polls released this year on X/Twitter, Trump is leading in the 2024 electoral competition, securing an average of 72% of votes as opposed to Biden’s 28%. Grabowicz and his team have introduced a website, socialpolls.org, which will be regularly updated to monitor such polls and rectify any partiality.
“Our research warns about the lack of openness in social media platforms, even concerning significant events like national elections,” states Grabowicz. “If this can occur in that scenario, it is highly likely that similar situations exist in numerous other contexts.”
Image Source: Thomas Dutour / Shutterstock
