Evolving Dynamics in Public Opinion: Polling Trust, Partisan Cooperation, and Election Skepticism
Chapter 1: Investigating Distrust in Public Opinion Polls: Demographic, Political Attribute, and Behavioral Correlates
The apparent failures of pre-election polls have cast doubt on Americans’ faith in public opinion research. In an expert survey of 163 AAPOR members, an overwhelming majority attributed a decline in polling trust to 2016 errors, while opinions were more divided on whether distrust depressed response rates in 2020. Reflecting these concerns, over 30% of experts reported omitting political questions from their instruments to avoid alienating participants. Although scholars last benchmarked national trust in polls and other polling related attitudes over a decade ago (Kim et al., 2011), contemporary evidence has been scarce. This paper updates our understanding by examining five data sources over multiple decades to chart aggregate trust in polling and identify predictors of skepticism. While aggregate trust has remained stable since the early 2010s, older individuals, Republican partisans, and conservatives consistently report lower trust. Additionally, I find no association between polling trust and whether citizens participated in the 2022 primary and general elections. These results underscore the value of routinely benchmarked attitudinal metrics in the public and reveal recent partisan cleavages in polling trust. Furthermore, I demonstrate that distrust in polling does not translate to depressed electoral participation, even amid heightened scrutiny towards polling by political elites and the media.
Chapter 3: Evaluating the Role of AI in Correcting Election Misinformation
Misinformation is widely considered to be one of the great modern threats to democratic governance and civil society. False beliefs about elections are particularly troubling, as they directly undermine the legitimacy of electoral winners and corrode public confidence in governing bodies. The efficacy of corrective interventions to address misinformation broadly has been well-documented, especially in the U.S. (Walter et al., 2020; Blair et al., 2024); however, there has been notably less research on correcting election misinformation (Nyhan, Stainfield and Weaver, 2024). What work has been done suggests that, while corrections are effective at reducing belief in the specific claim targeted (Clayton and Willer, 2023), the positive effects do not extend to increased confidence in American elections generally (Carey et al., 2024). This is unsurprising, given that corrections do not impact attitudinal or behavioral constraint except under idiosyncratic circumstances (Porter and Wood, 2024). Recent research, however, shows that brief interactions with AI-powered chatbots not only reduced belief in a single conspiracy theory but also diminished overall conspiratorial thinking among self-identified conspiracy theorists, with effects lasting up to two months (Costello, Pennycook and Rand, 2024). Using a two-wave survey experiment, this paper investigates the effect of supplementing standard corrections with AI chatbots on the beliefs and confidence of election skeptics, as well as the durability of these effects two weeks post-treatment.
Chapter 2: Vibe Check: How Political Context Shapes Partisan Survey Participation
Political polls and surveys have historically provided a critical window into public opinion. However, recent studies suggest that survey participation is increasingly shaped by political engagement and partisan asymmetries (Cavari and Freedman, 2023; Clinton, Lapinski and Trussler, 2022). While some scholars attribute these patterns to electoral context (Gelman et al., 2016; Borgschulte, Cho and Lubotsky, 2022) others find no systematic difference based on partisanship (Hopkins and Gorton, 2024). I propose that partisan differences in survey participation are not influenced by election-related factors alone (such as campaign activity or media coverage), but also follow a distinct time-based pattern that aligns with which party controls the presidency, moderated by the president's approval rating. When a Democrat is in office, Democratic respondents may be more likely to cooperate with surveys, while Republican respondents may disengage (and vice versa). Moreover, partisans may respond to shifts in presidential approval by opting in (or out) of surveys when the president has positive (or negative) approval among the general public. Leveraging over 100 waves of Pew’s American Trends Panel (2014–2023) and over 1,900 estimates of presidential approval from public opinion polls archived by the Roper Center at Cornell University, I anticipate evidence of differential partisan cooperation. Specifically, I expect heightened response rates among co-partisans of the sitting president, with pronounced increases following general elections and modest fluctuations during dips (and spikes) in presidential approval.