Using actual comments scraped from Facebook (~6M) and comments elicited from a representative sample of Americans, we show that self-selection of committed partisans exacerbates the toxicity of online discussion, and exposure to toxicity inspires others to write toxic comments themselves, thus creating a vicious cycle.
We suggest an alternative conceptualization of selective exposure: partisans select when to pay attention to politics, instead of which ideological sources to follow, such that they modify their political attentiveness in response to whether the flow of information is congenial to their party. We find support for our hypothesis in two studies.
We test popular conjectures about differences in the policy response to the current opioid epidemic and the crack scare of the 1980s–1990s using data on district-level drug-related deaths and legislation on illegal drugs in the House of Representatives. Our results suggest that the racial inequalities and double standards of drug policy still persist but in different form.
Using a difference-in-difference strategy that compares over-time belief changes of those interviewed before and after a sudden diffusion of rumor about Obama’s religion in 2008, we find that this event increased people’s belief that Obama is a Muslim by 4 to 8 percentage points.
Two experiments show that partisans are just as receptive to uncongenial information as congenial, absent partisan affect primes. But when randomly induced to feel adversarial, partisans become more dismissive of uncongenial information— and they end up disagreeing more, not less, after considering the same information. These results (1) establish the bounds of resistance to political persuasion; (2) identify partisan motivated reasoning more clearly than previous studies; (3) highlight the importance of the quality of elite-level political discourse in determining the quality of citizen-level opinion formation.
We examine how the American public—Republicans in particular—respond when provided with a relatively detailed causal explanation summarizing why scientists have concluded that human activities are responsible for climate change. Based on a well-powered survey experiment (N = 3007) assessing the effectiveness of detailed causal evidence vs. traditional consensus messaging, we found that both information treatments increase belief in anthropogenic climate change. We conclude that partisan resistance to science communication may not be the leading cause of the polarization of the American public’s beliefs and opinions on climate change.
Drawing on 1% random draws of 2015 and 2020 voter files that cover nearly all Americans and include information on partisan composition of each household, we examine whether Americans have indeed become more likely to live in politically segregated households in recent years. Our analyses indicate that, on average, household-level partisan homogeneity did not change. However, among young people, partisan homophily increased modestly—a pattern that was more salient among (1) those who lived in two-adult households, (2) those who were politically active, and (3) those who lived in politically homogeneous areas. These findings suggest that young Americans’ political polarization has continued to grow in recent years, albeit marginally.
We argue that presidential voting typically results in eventual disappointment—either because the preferred candidate loses or because the elected president’s big promises are under-delivered. Applying a regression discontinuity design to the ANES time-series data (1974 to 2008), we find that being just eligible to vote in a presidential election undercuts political trust 2 or 4 years down the road by several percentage points.
First Place Winner, the Annual Student Paper Competition, American Association for Public Opinion Research (DC Chapter)
We distinguish two dimensions of citizens’ political knowledge; information—the classic definition—and consideration of diverse perspectives, and argue even highly “informed” citizens in the traditional sense may fail to make “considered” decisions. A discriminant validity test establishes the lack of correspondence between information and consideration. We draw on imputation methods and a deliberation experiment with a representative sample to show that information and consideration may have different effects on collective opinions.
Top Student Paper Award and Travel Grant, the Political Communication Division, International Communication Association
What is the effect of an increase in digital communication on social movements, across democracies and non-democracies? In this study, we leverage rapid year-to-year changes in internet access at the country level to estimate the effects of an increase of internet access on political movements around the globe, and explore heterogeneous effects by political context. We use a regression discontinuity and a difference-in-difference designs to estimate the immediate and long-term effects of internet expansion. We find no evidence that a rapid expansion of the internet immediately affected political movements—a pattern that was consistent across democracies and non-democracies. We find, however, that internet access hampers political movements in the long run, particularly in established democracies. These findings highlight the importance of considering the over-time and context-dependent variations in the effects of the internet on political movements.
Project sketch: (1) A natural experiment randomly assigning people to respond a survey either just before or just after the Kavanaugh confirmation vote in 2018 to estimate the effects of his confirmation on partisan and gender gaps; (2) Survey experiments using counterfactual questioning to measure partisans’ reactions to the allegations made against Andrew Cuomo, Joe Biden and Eric Greitens (3) A survey experiment randomizing the partisanship of accused individuals and the credibility of allegation to confirm asymmetric reactions between Democrats and Republicans. (4) A survey experiment on how the public’s reactions to sexual harassment vary depending on the extent to which the accuser deviates from the ideal victim
We track the public’s attitudes toward the Supreme Court of the United States across three major events that have important implications for gender politics: the confirmation of Brett Kavanaugh, and the rulings on the abortion ban in Texas, and the upcoming decision on Roe vs. Wade.
Using a novel research design that randomizes the timing of survey invitation (before and after the Kavanaugh confirmation), I measure the effects of the confirmation on public opinion. I find that Kavanaugh’s confirmation did not increase the gap between women and men, but between Republican women and non-Republican women.
Featured in The Dartmouth (student newspaper at Dartmouth)