2020-2021 Spring Colloquia

ABSTRACT: Human-coded data is the basis for research and decisions in the social sciences, industry, government, and beyond. But coders often make mistakes, are biased, or are unmotivated, and such errors render standard coding methods unreliable and inefficient. In this paper, I explain how explicitly modeling coder mistakes helps to overcome errors in human coding — increasing the efficiency and reducing the bias of the final analysis. I introduce a new model, the biased-annotator competence estimation (BACE), as a default coding model for typical social science coding tasks. I prove the conditions for identification of the key parameters of interest for this newmodel and clarify the identification conditions for several models widely used across computer science and statistics. In simulations, I show that BACE serves as a viable default model for human coders. In an application, I show that, once corrected with BACE, there is 2-3 times more partisan polarization in the discussion of presidential candidates than would have be found using conventional hand-coding methods. I provide an easy-to-use R package that makes these models immediately applicable to coding tasks.

ABSTRACT: Democratisation literature mostly focuses on the elite’s decision to extend the franchise. But in many cases, current voters have to decide whether to grant the vote to a broader population. Little evidence exists on what factors drive the support among those who are already enfranchised. In this paper, I exploit the change in municipal Yes-vote shares among male voters for two Swiss national referenda on female suffrage between 1959 and 1971. I show that municipalities, which quasi-randomly introduced local female suffrage in between the two referenda, increased their support much more. This increase is driven by municipalities in which a majority of men was initially opposed to national suffrage. Conditioning on similar initial support, I further show that this difference cannot be explained by a “ceiling effect”. My findings can also not corroborate that the rise in support is driven by post-suffrage change in municipal party vote shares, expenditure, or cultural proxies, such as female labour market participation.

ABSTRACT: Abstract concepts like political ideology are hard to mentally represent, so some scholars have argued that we ground them in concrete concepts through metaphor. One prominent example is Lakoff’s ‘government as family’ theory, suggesting that our perspective on how government should function is based largely in our perspective on how families should function. Lakoff has suggested, for example, that a preference for nurturant (as opposed to disciplinarian) parenting style will yield different attitudes on issues such as welfare, prison sentences, and law enforcement. In this paper we explore the recent rise of helicopter vs. free range parenting styles though the lens of Lakoff’s model, finding evidence of a link between helicoptering and preference for paternalistic government policies, while free range parenting styles yield a preference for libertarian policies. This parenting metaphor explains more of the variance than ideology, party affiliation, and several other common predictors of political preference. We then explore the extent to which this metaphor extends beyond political settings, exploring the implications for business, education, health care and several other domains.

ABSTRACT: People are notorious for motivated reasoning, or the application of cognitive skills to reinforce and defend their attitudes. Motivated reasoning strategies include the disconfirmation bias, in which people strenuously search for flaws or other reasons to reject evidence that challenges their attitudes. What happens, however, when the evidence challenging our political attitudes is so compelling that it cannot simply be denied or ignored? Enter framing. Most of us know of framing as a tool of political communicators. We argue that people use framing to uphold political attitudes in the face of challenging evidence. Specifically, people reframe the issue so as to diminish the importance of beliefs that have been challenged, thereby preserving attitudes linked to those beliefs. We present evidence from two experiments that is consistent with the reframing hypothesis. Honestly, it really is.

ABSTRACT: Research on personality and political preferences generally assumes unidirectional causal influence of the former on the latter. However, there are reasons to believe that citizens might adopt what they perceive as politically congruent psychological attributes, or at least be motivated to view themselves as having these attributes. We test this hypothesis in a series of studies. Results of preregistered panel analyses in three countries suggest reciprocal causal influences between self-reported personality traits and political preferences. In two two-wave survey experiments, a subtle political prime at the beginning of a survey resulted in self-reported personality traits that were more aligned with political preferences gauged in a previous assessment. We discuss how concurrent assessment within the context of a political survey might overestimate the causal influence of personality traits on political preferences, and how political polarization might be exacerbated by political opponents adopting different personality characteristics or self-perceptions thereof.

ABSTRACT: Survey research suggests that Democrats and Republicans disagree over matters of fact, but little evidence speaks to whether these disagreements reflect ignorance of the truth or misinformed belief in falsehoods. Based on two key findings, I argue that measured partisan belief differences primarily reflect differences in knowledge and ignorance, not belief in falsehoods. First, respondents often state considerable uncertainty about incorrect answers. Second, conditional on stated certainty, incorrect answers tend to be temporally unstable. Despite the overwhelming tendency of incorrect answers to represent uncertain guesses, measuring partisan belief differences in a manner that allows individuals to express uncertainty about their answers does not uniformly shrink measured belief differences. This is because correct guessing creates an illusion of knowledge of inconvenient truths, countering the influence of generalized uncertainty. In some cases, partisan belief differences increase when individual-level uncertainty is accounted for. These findings suggest that prevailing practices in survey research warp portraits of the public’s factual beliefs, suggesting widespread misperceptions in areas where few exist while dulling observers’ sense of where partisan divides are largest and deepest. This encourages observers of politics to over-apply misinformation-based explanations at the expense of ignorance and selective exposure. The evidence on the measurement properties of incorrect answers calls for reinterpretation of a wide range of research that purports to examine the prevalence, predictors, and consequences of misperceptions and misinformed beliefs.

ABSTRACT: In 2020, the United States faced multiple simultaneous crises such as wildfires and COVID-19. In two national surveys we investigate how risk perceptions, efficacy perceptions, social influence, climate change perceptions, subjective attributions, and personal experience are associates with mitigation/adaptation intentions and behaviors regarding wildfire smoke and COVID-19. Today, we will present three sets of preliminary results using the national surveys data. The first examines the relationships of perceived social norms, social support, threat, and efficacy on intended mitigation behaviors, such as mask wearing. Our initial findings suggest that perceived social norms may have a direct effect on mitigation behavioral intentions and also that in some settings, social support may amplify perceived self-efficacy, which is associated with greater behavioral intentions. The second examines perceived threat, perceived efficacy, experience, and place attachment as predictors of perceived likelihood of future wildfire-associated migration. Initial results indicate that prior smoke exposure is associated with lower perceived threat of wildfire and wildfire smoke, while greater perceived threat and lower place attachment are associated with greater reported likelihood of future wildfire-associated migration. Finally, the third examines the relationships between climate change beliefs, experience, subjective attributions regarding wildfires and COVID-19, and the extent to which subjective attributions are related to pro-environmental individual and collective behavioral intentions. Our initial finding suggest that climate change perceptions and personal experience are associated with subjective attributions, and that these attributions are associated with pro-environmental behavioral intentions.

ABSTRACT: How can the fate of political campaigns be forecast? Some theorists claim that rapid implicit responses to political candidates determine eventual campaign success, while other argue that the changing course of events over time drive campaign viability. In this research, we sought to determine whether neural activity in a group of subjects (n=41) could forecast the fates of the political campaigns of 15 nominees over a year during the Democratic primary of 2019-2020. Consistent with a “partial scaling” neuroforecasting account, we predicted that while individuals’ brain activity might predict their endorsement of candidates, group brain activity would also forecast the eventual survival of candidates’ campaigns. As predicted, group MPFC activity forecast political campaign survival, but behavior (e.g., endorsements) did not. Moreover, the combination of group brain activity and behavior forecast campaign survival better than contemporaneous polling. Theoretically, these findings support a partial scaling account of neuroforecasting by demonstrating that brain activity can forecast significant political outcomes. Practically, the findings hold implications for enhancing the efficiency and impact of political campaigns by suggesting that early subjective responses to candidate attributes may contribute to campaign survival.

ABSTRACT: The American National Election Studies (ANES), the gold standard for nationally representative opinion surveys of the American electorate, has employed a core set of questions since 1948 so as to track trends over time and to compare elections with one another. Since then, a robust literature on questionnaire design has come into being, highlighting various opportunities for improving suboptimal survey designs. In 2008, a series of question experiments designed to test alternative question wordings that the survey research literature suggests are improvements were spliced into the ANES Time Series Study. This paper analyzes these questions experiments for the first time, assessing concurrent and predictive validity of the “improved” questions using OLS regression, structural , and machine learning models. In total, I find that eleven out of sixteen sets of improved questions yield higher validity across a number of criteria, suggesting that the ANES and other organizations interested in survey research should implement these or similar question versions in future studies in order to more accurately measure American public opinion.