Using our interactive tool to examine social posts, we found numerous false claims of election fraud and differences in the online conversations in English versus Spanish.
The Midterm Monitor is a joint project of the Alliance for Securing Democracy at the German Marshall Fund and the Brennan Center. Launched in September 2022, it gathers data from accounts affiliated with candidates for governor, secretary of state, the House, and the Senate. In addition, it captures data from the accounts of influential national media outlets and pundits, the most followed local media outlets in 10 battleground states, non-English-language media outlets in the United States, and state media and diplomats associated with the Chinese, Iranian, and Russian governments.
We are using the Midterm Monitor to conduct and publish analyses to help understand the online messaging about the 2022 midterm elections. Last week we published some of our findings. Here are some additional trends we are seeing:
- Secretary of state candidates are perpetuating false narratives about the election online. In comparison to September, during the first half of October we saw a decline in the incidence of false narratives about the election on Facebook and Instagram, although they remained at a high level. On Twitter the incidence of false narratives has increased over the past two reporting periods.
- As we get closer to Election Day, the discussion of fraud as a topic among all candidates is increasing. This discussion includes comments made by individuals speaking out against election denialism, as well as fraudulent claims by election deniers.
- There are differences in the online conversations in English versus Spanish, including which candidates are posting in these languages and what topics they are discussing.
Election denialism remains central in the online conversation among secretary of state candidates.
In comparison to the last period, we saw a decline in the incidence of false narratives on Facebook and Instagram. The incidence of false narratives for Twitter has increased over the past two reporting periods.
- Of 356 English language Tweets from secretary of state candidates between October 1 and October 18, 25 percent, perpetuate a false narrative.
- Of 234 Facebook posts by secretary of state candidates between October 1 and October 18, 21 percent, perpetuate a false narrative.
- Of 68 posts shared by secretary of state candidates on Instagram between September 28 and October 12, six, 9 percent, perpetuate a false narrative.1
Fraud is becoming an increasing topic of conversation as the election nears, particularly for secretary of state candidates.
Fraud in American elections is exceptionally rare, but as we get closer to the election, online discussion of the topic is increasing. We used a number of different methodologies to explore the discussion of fraud. Topic modeling of the discussion of fraud by all candidates on Twitter between January 1 and October 23 (at two-week increments) shows that discussion of fraud has gradually risen over the course of this year.2
Our previous analysis showed that fraud was a central topic of conversation among secretary of state candidates. Our additional analysis further demonstrates that it is not only a central topic, but one that is growing in prominence in the candidate Twitter conversation about the elections.
By the same token, we utilized NYU Ad Observatory and Meta Library to research advertisement spending across Facebook and Instagram, which are owned by Meta. A total of $177,594 was spent on ads containing the word “fraud” since January 1, but the largest sum ($18,194) in nearly six months was spent during week of October 25.
Our analysis further revealed that some candidates may also be more likely to discuss fraud than others. According to NYU Ad observatory data and categorizations, ads featuring “stop the fraud” were mostly funded by right leaning spenders. During the month of October, secretary of state candidates were also more likely to use the word “fraud” in Twitter or Facebook posts, compared to other candidates.
There are important differences between Spanish and English online conversations about the election.
Topic modeling of Spanish-language posts about the election between mid-August and mid-October showed that Spanish-language media was more likely to talk about foreign elections in their coverage than English-language media outlets.3 In Spanish, there was more discussion of topics such as voter fraud in Brazil, Italy’s new prime minister, and other foreign elections, with many posts focusing in particular on election fraud in other countries.
Finally, we found that some but not all candidates are tweeting in Spanish. Approximately 18 percent of candidates tweeted in Spanish at least once during September or October. Nearly twice as many Democratic candidates tweeted in Spanish compared to Republican candidates. The candidates tweeting most in Spanish were from Texas, California, and Florida, which all have large Spanish speaking communities. The centrality of Spanish, particularly in these three major states, underscores the importance of tracking the online conversation in Spanish.
The views expressed in GMF publications and commentary are the views of the author alone.
- A team of Brennan Center analysts hand coded posts by secretary of state candidates on Twitter, Facebook, and Instagram for the month of September. The analysts coded for 37 different false narratives related to election administration. The list was created by Brennan Center researchers. The false narratives are based on key narratives perpetuated ahead of and after the 2020 election.
- This analysis looked at the discussion of fraud as a topic rather than only fraudulent statements. This analysis does not differentiate between posts that perpetuate fraudulent narratives and those that do not.
- Topic modeling is a machine learning approach. It uses statistics and natural language processing to identify and cluster topics within text.