Samuel Frantz, Thessalia Merivaki and Michael Baldassaro. "When Misinformation Turns Toxic: Using Machine Learning to Measure Toxicity in Social Media."
The 2020 U.S. election illustrated the capacity of online misinformation to fuel offline violence, although research has not yet established a causal link between the two. In this paper, we argue that for this relationship to hold, misinformation will need to reach critical levels of toxicity. Drawing upon public posts on Facebook prior to the 2020 election, we show how the spread of misinformation narratives led to an increase in toxic speech online. We identify when misinformation narratives took place within groups and individuals, and then use natural language processing tools to assign a toxicity score to content before and after the introduction of the narrative. We validate the causal link between misinformation and toxicity through a case study of conspiracy theories surrounding voting machine manufacturer Dominion Voting Systems, where content propagated online lead to threats of violence offline. Methodologically, the paper is among the first to apply the Google Perspective API to the study of politics. This tool, developed to assist with online content moderation, utilizes advances in machine learning to classify speech as potentially toxic. Our research has important implications about misinformation and its connection to the degradation of democratic discourse and electoral violence.
Merivaki Thessalia, Christopher Mann and Ioannis Ziogas. " 'Click Here to Register to Vote,' and Then What? Assessing the Impact of Voter Registration Procedures on Prospective Voters."
Researchers find that voter participation increases when states remove institutional barriers, such as voter registration deadlines. However, turnout has not significantly increased over time, raising the question of whether election reforms yield the intended effects. We identify one caveat in extant research of voter participation, namely separating eligible-to-vote and interested-in-voting individuals from eligible-to-vote but not interested-in-voting individuals. We argue that even when prospective voters actively attempt to participate, there are structural challenges which might not allow them to cast a vote on Election Day. We focus on the process of registering to vote online, given that more states offer Online Voter Registration (OVR) and/or make available registration forms on state election websites, and show that there is notable variation in OVR features, successful voter registration, and turnout. We track eligible U.S. citizens who initiated the registration process – “ intended registrants”- using the voter registration portal at Vote.Org, and assess how the availability of OVR, registration deadlines, and various OVR features affected voter participation in 2018. We test the robustness of our findings by replicating the analysis with data from the 2016 election. These novel datasets allow us to uncover nuances in the voter registration process that have not been studied before and demonstrate persistent barriers to voting after controlling for interest in elections.
Merivaki Thessalia and Enrijeta Shino. "Measuring the Activity of Voter Registration Drives in the Midst of a Pandemic." (Working paper for SPSA 2021)
The role of voter registration drives in registering eligible citizens to vote and mobilizing new voters is heavily discussed in the literature of voter turnout. Yet, the impact of voter registration drives on turnout is understood in a broad manner, as if all groups operate under the same structures or capacity. The COVID-19 epidemic has made face-to-face outreach challenging, thus directly impacting the ability of advocacy groups, campaigns, and individuals to reach prospective voters. How do registration drives operate in this environment, and how does their activity impact voter registration rates at the local level? In this paper, I attempt to measure the density of Third-Party Voter Registration Organizations (3PVOS) across Florida’s 67 counties during the 2020 election cycle and identify patterns in how geographically spread they are. I use data on registered 3PVROs, as well as voter registration and turnout statistics reported by the Florida Division of Elections to assess whether new voter registration rates fluctuate as the density of these 3PVROs changes at the county level.
Endicot W. Travis, Thessalia Merivaki and Julie Wronski. "Can Voters Tell When Costs of Voting are High? The Voter Experience In Mississippi." (Under Review in Journal of Race, Ethnicity and Politics)
Recent studies of voter turnout suggest that restrictions in voter registration access and modernization, as well as restrictions in voting convenience increase the costs of voting. Mississippi takes the lead by being one of the states with the highest costs of voting (Li et al. 2018). As a traditionally Republican, non-battleground state with consistently low turnout, the media usually pays little attention to Mississippi, and this state is often omitted from traditional exit polls in presidential general elections. Compounding this, the voter experience in Mississippi has not been adequately studied by academics. Thus, it is unclear how exactly Mississippi's controversial voter laws (which are currently subject to a federal lawsuit), and the racially polarized nature of partisanship within the state, impact individual voting behavior.
We address this lacuna by trying to shed light onto the attitudes and voting choices of Mississippians. In this study, we gauge the experiences of approximately 400 voters across counties that were predominantly rural, or home to a major university, using student-conducted exit polls in the 2018 midterm and 2019 gubernatorial elections. In 2018 we find that non-Black respondents with more trust in the government, are more likely to see voting as an effective way for people to have a say in government. In both years we find that White respondents with high trust in government were more confident in their vote being counted accurately. Compared to White voters in 2019, Black respondents were less confident that their ballots would be counted accurately, and that voting is an effective way to have a say in government.