“Who Cares? Measuring Differences in Preference Intensity” (with Daniel L.Chen and Karine Van Der Straeten). [Under Review]
How well do existing survey methods capture differences in preference intensity? To answer this question, we measure people’s preferences with one of three methods: (1) Likert items, (2) Likert items followed by issue importance items and (3) Quadratic Voting for Survey Research (QVSR). Using real-world decisions as a proxy for preference intensity, we find that issue importance items and QVSR outperform traditional Likert items for identifying who cares, with QVSR demonstrating a noticeable advantage. In light of these results, we revisit a longstanding debate on the material determinants and self-interested origins of policy preferences. Conclusions, we show, can vary with the measurement tool. We discuss implications for empirical research involving subjective survey data.
“Willingness to Say? Optimal Survey Design for Prediction” (with Daniel L.Chen, Ritesh Das, and Karine Van Der Straeten). [Under Review]
Survey design often approximates a prediction problem: the goal is to select instruments that best predict an unobserved construct or future outcome. We demonstrate how advances in machine learning techniques can help choose among competing instruments. First, we randomly assign respondents to one of four survey instruments to predict a behavior defined by our validation strategy. Next, we assess the optimal instrument in two stages. A machine learning model first predicts the behavior using individual covariates and survey responses. Then, using doubly robust welfare maximization and prediction error from the first stage, we learn the optimal survey method and examine how it varies across education levels.
“Why Some Care More About Free Riding Than Others and Why It Matters.” [Under Review]
People support policies that increase their own expected income. They also support policies that move the status quo closer to what is prescribed by agreed-upon norms of fairness. How do these two motives combine? In most circumstances, I argue, people reason as moral agents trying to do the “fair” thing. Only when status quo changing policies have large and certain material consequences will they deviate from saying what is fair and choose to express a self- serving position instead. I apply this simple framework to a form of fairness reasoning that ties social policy preferences to beliefs about the prevalence of free riders among net beneficiaries of social spending. I show that income level and the institutional context affect the extent to which pocketbook concerns overrun free riding ones. I flesh out the implications for the politics of social policy reform in mature welfare states.
“Much Ado About Debt: Understanding How People Reason About Debt (Un)Sustainability.” (with Bjorn Bremer, Lisanne de Blok and Catherine de Vries)
Is there an electoral basis for debt consolidation? Evidence showing higher than expected demand for fiscal discipline suggests there is. This paper uses novel British survey data to probe this possibility further. First, we examine the existence of a pro-consolidation subconstituency, that is, an influential subset of voters who prioritize fiscal discipline over other issues. Second, we turn to persuasion bias and investigate whether voters tend to resist arguments describing high debt levels as sustainable while more willingly accepting arguments describing them as unsustainable. Third, we test for the existence of mental models that exist independently of partisan messaging and predispose voters to favor fiscal discipline. We find no evidence that British voters tilt the balance in favor of fiscal consolidation in these ways, and discuss implications for future research on fiscal politics in Western democracies.
“Explaining Differences in Free Riding Beliefs: A Review and Preliminary Theory.”
People who oppose generous social benefits for the poor and the unemployed often believe that recipients are free riding, i.e., abusing society’s generosity by failing to act to improve their plight. People who support more generous social benefits express the opposite concern: the moral wrong is on the side of society, which does too little to help people who cannot be blamed for their economic conditions. What explains these differences in welfare attitudes and free riding beliefs? To answer this question, this paper focuses on the well-documented, yet puzzling, correlation between 1) free riding beliefs on the one hand and 2) liberal-authoritarian values on the other. Underpin- ning this correlation, I hypothesize, is a disagreement over how to best address social dilemmas, i.e. how to maximize pro-social behavior and minimize free riding. I provide preliminary evidence for this line of inquiry and conclude by discussing implications for future research.
“Left Behind by Linear Regression? Insights from Machine Learning Applied to Brexit.” (with Sabina Tomkins)
Who in Britain voted to leave the European Union and why? In search of answers, scholars use linear or logistic regressions applied to observational data to identify individual and contextual characteristics associated with higher support for Brexit. A common conclusion is that the Brexit vote is the symptom of a new political cleavage pitting those left behind by free trade and globalization against university-educated workers who benefit from a shift to a globalized high-skill service economy. This paper complements the model-based approach commonly found in the Brexit literature with machine learning’s emphasis on predictive modeling. We find that for people with individual Leave features and contextual Remain features, a flexible machine learning approach improves performance by 5%. For individuals with the opposite mix (individual Remain and contextual Leave) the performance improves by 11%. Because these groups represent a minority of the sample (40%), overall improvement is more limited, explaining why a parametric linear or logistic model using only individual features is commonly deemed adequate. Using our preferred model, we find that 40% of the test sample aligns with the maximalist left-behind interpretation of Brexit pitting places left-behind by economic and social changes against prosperous ones who benefit from them. In total 20% behave exactly the opposite of what this theory predicts. We discuss implications for assessing the importance of contextual variables relative to individual ones and for theory development more generally.
Charlotte Cavaille 