Abstract District-based elections are a central feature of local governance throughout the United States. Prior work has explored whether district-based elections impact racial/ethnic descriptive representation in local office; much less is known about the impacts of local district-based elections on other dimensions of representation. We consider another such dimension: socioeconomic class. To explore how district-based elections shape the composition of locally elected officials on class dimensions, we focus on city councils and study the dramatic shift towards district-based elections in California in the 2010s. We construct a statewide mapping of newly drawn council districts; we also draw on rich and partially hand-collected data on council candidates and members. We find that district-based elections increase the share of candidates and council members from lower-income and higher renter share neighborhoods, and lead to fewer members with business backgrounds.

Abstract Inter-party communication is crucial in representative democracies, facilitating information exchange and dialogue among political parties. Despite its importance, research on this topic remains limited due to lacking conceptual clarity and challenges in large-scale measurement. This article offers a comprehensive definition of inter-party communication as public communication by parties about others, with a positive, neutral, or negative stance, focusing on collaboration, policy, or personal issues. To effectively measure this phenomenon, we introduce a novel transformer-based approach capable of automatically classifying large volumes of text. Case studies on coalition signals in Germany and negative campaigning in Austria demonstrate its effectiveness. The study deepens our understanding of party competition, advances methods of automated text classification, and enables new research on political communication.

Abstract Why would a strong authoritarian state choose not to enforce its own policy? We extend the theory of forbearance to autocracies, highlighting its distinct incentives and characteristics. Using China’s social insurance policies as a case study, we argue that promotion-driven local officials under intense interjurisdictional competition allow firms to evade payroll taxes to boost economic performance and advance their careers. This effect is most significant among domestic private firms and foreign firms. We conduct one of the first systematic analyses of firm-level social insurance contributions in an authoritarian context, supplemented by individual-level survey data. Our findings show that bureaucratic forbearance of China’s social insurance policies has a pro-business bias, undermining the policies originally designed to address inequalities during market reforms.

Abstract Recent scholarship has emphasized methodological innovations to mitigate preference falsification in public opinion data, yet systematic scrutiny of bias in regression analyses remains limited. Drawing on analyses of political trust in China, we offer three key insights. First, determining the direction of social desirability bias in regression estimates—whether over- or underestimation—is challenging ex ante . Second, analyses of two nationally representative Chinese surveys, one incorporating a list experiment, cast doubt on the purported positive effect of social welfare expansion on political trust. Extending beyond social welfare and the Chinese case, we find similar biases when regressions rely on direct questions. Third, we show that certain identification strategies can partially mitigate regression bias when direct questions are unavoidable.

Abstract Understanding how political information is transmitted requires tools that can reliably and scalably capture complex signals in text. While existing studies highlight interest groups as strategic information providers, empirical analysis has been constrained by reliance on expert annotation. Using policy documents released by interest groups, this study shows that fine-tuned large language models (LLMs) outperform lightly trained workers, crowdworkers, and zero-shot LLMs in distinguishing two difficult-to-separate categories: informative signals that help improve political decision-making and associative signals that shape preferences but lack substantive relevance. We further demonstrate that the classifier generalizes out of distribution across two applications. Although the empirical setting is domain-specific, the approach offers a scalable method for expert-driven text coding applicable to other areas of political inquiry.

Abstract Social scientists often compare survey responses before and after important events to test how those events impact respondent beliefs, attitudes, and preferences. This article offers a formal analysis of such pre-event/post-event survey comparisons, including designs that seek to reduce bias using quota sampling, rolling cross-sections, and panels. Our analysis distinguishes major sources of bias and clarifies the comparative strengths and weaknesses of each approach. We then introduce a modified panel design—the dual randomized survey—to reduce bias in cases where asking respondents to complete the same survey twice could impact their Wave 2 responses. Our formalization of bias and novel research design improve scholars’ ability to study the causal impact of events through surveys.

Abstract Do anti-elitist parties behave differently in parliament than other parties? Existing evidence is inconclusive: some studies suggest that anti-elitist parties do not show a shared voting pattern as this is mainly structured by their left- or right-wing ideology. Others suggest that these parties vote against legislation more often. In order to address this question, we develop a new method that allows one to look at different explanations of voting concurrently while also taking into account characteristics of the vote. We find that anti-elitist parties do vote in a similar way and different from other parties, but only on legislative votes. As such, we present a major step forward in our understanding of and methodological approach to parliamentary voting behavior.

Abstract Is an autocracy’s official rhetoric a reliable proxy for forecasting military escalation? While the conventional hands-tying mechanism argues that official rhetoric binds leaders to stated positions and limits their ability to back down, recent scholarship on bluster suggests that autocracies may employ hawkish rhetoric to justify de-escalation ex-post. This study evaluates these competing perspectives by analyzing China’s official rhetoric and military behavior in the Taiwan Strait from 2016 to 2022. Employing a word-embedding approach, I construct an original Chinese-language lexicon capturing implicit threats from over two million state-media articles. I show that increases in China’s implicit threats toward Taiwan are associated with a higher likelihood of military escalation, implying that official rhetoric conveys predictive information rather than mere cheap talk.

Abstract Measuring party loyalty (and party effects, more broadly) in roll call voting has long been a contentious matter in the study of legislative behavior. While techniques for measurement in this arena are numerous, most of them suffer from a fatal flaw: they improperly (or insufficiently) separate measures of preferences from party effects. A significant part of this measurement challenge is the identification of (a) which roll calls leaders care about and (b) in which direction they desire their members to vote. In this paper, we use a novel dataset of party leader speeches from the 101st to 113th Congresses to develop a model of party loyalty and measure loyalty across members and time. Unlike existing techniques, we allow for party influence to vary across legislators and time. Additionally, our model provides estimates of legislator ideology and party loyalty disentangled from one another. Using these estimates, we explore the dynamics of loyalty in the contemporary Congress and unearth findings quite different from extant measures in the literature.

Abstract Estimates of sensitive questions from list experiments are often much less precise than desired. We address this well-known inefficiency problem by presenting an informative Bayesian approach that combines indirect measures with prior information. Specifying informed priors amounts to a principled combination of information that increases the efficiency of model estimates. This framework generalizes a range of different modeling approaches for list experiments, such as the inclusion of direct items, auxiliary information, the double list experiment, and the combination of list experiments with other indirect questioning techniques. As we demonstrate in real-world examples from political science, the informative Bayesian approach not only improves the utility but also changes the substantive implications drawn from list experiments.