Do group identities rooted in salient social and political characteristics foster homophily and social distance? If so, which identities shape social preferences more strongly, and under what conditions? While existing literature on affective polarization and inter-group relations privileges particular social and political identities, we know less about the dynamics that unfold when multiple identities are considered simultaneously. Using conjoint experiments in five multiparty systems (Argentina, Chile, Italy, Portugal, and Spain), we find that political identities are generally more influential than social ones among individuals with partisan affinities. This is especially true among left-wing individuals, who display greater social distance to radical right supporters. Our findings suggest that social distance is rooted in political attachments, which may channel salient social divisions.

Coalition parties face a unity-differentiation dilemma: they must uphold compromise to govern effectively, while maintaining distinct identities to maximize electoral success. Existing measures of coalition differentiation almost exclusively rely on static proxies to infer behavior, and the few dynamic measures often only cover a limited sample of governments, limiting comparative analysis. This article introduces a comparative, dynamic measure of rhetorical differentiation based on supervised classification of legislative speeches. Higher classification accuracy indicates greater rhetorical distinctiveness between coalition partners, while misclassification indicates greater rhetorical overlap. The measure provides monthly party-level scores for close to 50 coalition governments, across six European democracies, and thereby enables the dynamic analysis of coalition party differentiation in a public-facing arena, for a sizable comparative sample.

Citizens rely on information to evaluate the government, but their sources are limited in authoritarian regimes. This study shows that citizens can learn from extraordinary events, such as public safety incidents, which are socially salient, highly visible, and clearly attributable. When the state bears causal responsibility for a safety failure, political trust declines under both a controversial and an uncontroversial response. Using survey data that cover three national crises in contemporary China, the analysis compares respondents in the weeks immediately before and after the incidents. Trust in government declined following the 2011 bullet train crash and the 2018 vaccine scandal, but not following the 2008 Wenchuan earthquake. Further analysis finds no increase in perceived corruption, while perceptions of free expression declined after the train crash but not after the earthquake, consistent with dissatisfaction over information control. The findings have implications for authoritarian accountability and regime stability.

American political polarization has mapped onto educational pedigree, with credentialed elites increasingly supporting Democrats. We examine the implications of this trend in the most powerful office in American politics: the White House. From 1995 to 2024, we show Democratic presidents employed staffs with higher educational attainment, and that Republicans were more likely to attend public institutions, those without R1 status, and those ranked below the top 50%. We then show that universities whose alumni disproportionately went on to work for Democratic presidents were more likely to face investigations, enforcement actions, and funding cuts after President Trump took office in 2025. Our findings suggest educational pedigree among staff who operate key political institutions warrant further investigation.

How might changes in the credibility of one alliance relate to the credibility of other alliance commitments? Debate between competing views on this question spans decades in part because systematic tests are difficult. We contribute systematic evidence with a new measure of credibility in U.S. alliances and novel technique to estimate interdependence. Our measure draws on expert ratings of allied beliefs over U.S. credibility and our analysis employs Gaussian process regression to analyze credibility. We find strong correlations in allied credibility perceptions within regions and little correlation across regions. Supplementary analyses suggest that these correlations reflect common unobservables and a small impact of shared perception of past U.S. behavior.

Why are women underrepresented in civil service management across rich democracies? We argue that government ideology matters for this question, and that women bureaucrats’ careers stagnate under right-wing governments, since they do not suppress gender-inequality-reinforcing promotion patterns. We examine this using administrative occupation data on all Danish central government employees, observed quarterly from 1998 to 2019. The evidence is consistent with our argument: Under left-wing governments, men and women in the civil service progress similarly, while women progress more slowly than men under right-wing governments. The gender gap in career progression widens at the higher echelons of the bureaucracy, where political attention is greater, and this gap is not reduced by the presence of women ministers.

Elite cues signal citizens how to interpret political events through a partisan lens. Do these effects occur in real time? We exploit President Trump’s unexpected announcement of US military strikes against Iran during a nationally representative survey to estimate the causal effect of elite cues on public opinion. Leveraging minute-level variation in exposure, we find that partisan divergence emerges almost immediately: support for military intervention increases among Republicans but declines among Democrats. These large effects remain robust across 28,672 model specifications and 57,344 partisan-specific treatment effects. Our quasi-experimental evidence demonstrates that elite cues can polarize public opinion within minutes, even before media framing and partisan commentary have diffused, highlighting the speed of partisan opinion formation and its implications for democratic accountability.

Do larger polities hinder gender parity in elected office? We theorize that when populations grow faster than elected assemblies, a “seat congestion” effect intensifies competition for nominations, enabling men candidates to crowd out women. We test this argument using two nationwide natural experiments: Swedish municipal amalgamations in 1952 and 1973 that abruptly increased electorates. We track the effects in newly digitized historical panel data. Where electorate-per-seat ratios rose, growth in women’s representation slowed. Decomposition analyses show that expanding councils only partially offsets congestion. By identifying polity size as a systematic brake on descriptive representation, the study adds a demographic dimension to debates on institutional design and points to an overlooked cost of increasing jurisdictional size.

Felon disenfranchisement laws can alter the composition of the electorate. Accordingly, we argue that in states with such laws, elected prosecutors may be tempted to assign felon status strategically. We test our argument using criminal case data from Virginia and leveraging a sudden policy change—unexpectedly announced by the Governor on April 22, 2016—enfranchising all felons upon completion of their prison or probation sentence. Using well-known racial differences in partisanship as a measure of likely political alignment, we show that the announcement corresponded to an increase in criminal charge severity and sentence length for defendants who were politically unaligned with their prosecutors, and a decrease for aligned defendants. Evidence suggests that these changes may have been partially motivated by partisan concerns in the lead-up to the 2016 federal elections.

Demand for clientelism in low-income countries distorts political incentives and constrains growth. Informal workers are seen as a strong clientelist constituency because they rely on forbearance and targeted benefits, yet they also constitute an electoral bloc that could coordinate in support of beneficial programmatic policies. We examine whether—and how—such demand for clientelism can be attenuated. We implement an information experiment with firm owners ahead of Senegal’s 2022 local elections, during which an upstart political faction campaigned on programmatic reform. Reminding informal workers of their coordination capacity reduced engagement with dominant clientelistic networks. This shift did not translate into support for programmatic alternatives but into political disengagement. Demand for clientelism can be mitigated, but programmatic mobilization remains a challenge.

Do everyday encounters with foreign brands shape political attitudes? We argue that brand origin cues activate national identities, producing different responses among nationalists and cosmopolitans. In an online shopping experiment, U.S. participants were randomly assigned to compare American (control), European, or Asian (treatment) sport utility vehicles without explicit mention of brand origin. Exposure to foreign brands increased opposition to trade and immigration and support for Donald Trump among nationalists. Cosmopolitans generally showed little change. These effects are not reducible to partisanship, revealing within-coalition divides among Republicans and among Democrats and Independents. Our findings show that subtle consumer experiences spill over into politics, creating a bottom-up channel of polarization that links nationalism and populism to support for neomercantilist and nativist policies.

How we measure racial bias matters. Previous literature demonstrates that direct questioning leads to unreliable responses. Scholars have thus developed a number of techniques to uncover racial bias without cuing a defensive response. I investigate the performance of text- and image-based factorial experiments in the case of racial motivation for opposition to new housing development relative to a more traditional list experiment. My results demonstrate that these often-used methods yield substantially different conclusions about the importance of racial biases in opposition to new housing construction. While the factorial experiments suggest little evidence of racial bias, the more traditional list experiment reveals that the potential for racial change in their neighborhood is a concern for a meaningful share of respondents.

Contemporary democracies face growing public disaffection and mistrust. Governments have tried to revert these trends by promoting participatory governance. Prior studies established how deliberation can empower citizens. However, whether these effects hold in real-life initiatives initiated by partisan governments remains unclear. We test the ability of citizens’ assemblies to foster democratic attitudes in a natural experiment leveraging the random selection of participants to Lisbon’s citizens’ assemblies Participation led to an increase in issue knowledge and salience, external political efficacy and institutional trust. However, it did not impact internal political efficacy nor helped citizens better recognize the challenges of collective decision-making. The results hold for two assembly editions and reveal the potential of government-led participatory initiatives to reconnect citizens with democracy.

Does a politician’s religion influence their policymaking? While existing research has examined how gender, class, and ethnicity shape political decisions, the role of religiosity remains understudied. This paper investigates how religious values influence debt repayment decisions among local politicians. I develop a novel measure of individual religiosity, based on names matching local patron saints, and apply it to 62,908 candidates in Italian municipal elections. Using a regression discontinuity design, I find that narrowly elected religious Catholic mayors are more likely to increase debt repayment. This effect likely reflects two mechanisms: Catholic moral teachings, which stress the obligation to repay debt, and risk aversion rooted in religion. Overall, these findings demonstrate that religion meaningfully shapes the fiscal behavior of elected officials.

Many studies on candidate choice and other political preferences suggest that the effect of one attribute may mask the effect of another attribute, which the respondent has no information about. If, for instance, respondents lack information about party membership, they may infer that female candidates are Democrats, which would lead to biased estimates of the pure effect of candidate gender. We propose an alternative explanation—namely that respondents have lexicographic preferences—and we present an empirical test using conjoint experiments to distinguish between what we call pure masking and lexicographic preferences. We illustrate our method using data on the effect of gender, race, and party in candidate choice and replicate the results in a preregistered conjoint experiment. We offer recommendations for researchers studying candidate and political choices.

Political science increasingly uses text classification to gauge subtle concepts such as toxicity or anger. Traditional methods treat words in isolation, overlooking the contextual dynamics where meaning resides. Transformer-based language models address these limitations but remain underutilized, especially among applied scholars, partly due to misconceptions about their mechanisms and computational requirements. This article bridges this gap by offering an accessible explication of the pretrain-finetune paradigm, focusing on underlying mechanisms and their potential to improve political text analysis by harnessing models trained on extensive datasets while requiring modest labeled data and computational resources. I demonstrate the approach by identifying toxic language in conversation threads following U.S. Senators’ tweets and provide an online tutorial to support nontechnical scholars in adopting the methodology.

Preference intensity may play an important role in shaping mass political behavior, yet existing empirical scholarship on representation typically relies on survey measures that capture public opinion in terms of preference directionality without intensity. We use a variety of measures—not only the standard Likert and “most important policy” items but also the quadratic voting method—to reassess responsiveness and congruence of a diverse set of American state policies that vary in issue areas and salience. Across multiple novel measures of individual policy preferences, we consistently find that state-level preference intensity does not reliably predict actual policy outcomes. We contribute new evidence that the lack of policy representation in the American states remains—even if preference intensity is systematically integrated into the empirical analysis.

Vague legislative language can facilitate law adoption and implementation but reduce policy effectiveness and application uniformity. So, when do legislators employ it? Building on policymaking and responsiveness theories, we argue that legislators use vague language to overcome institutional conflict and public opinion constraints but avoid it when they seek to assert their positions in legislative output and policy outcomes. Empirically, we study the conditions under which the European Parliament (EP) introduces vague amendments to European Union (EU) legislative proposals using dictionary-based content analysis of EU legislative negotiations between 2009 and 2019. The results show that EP sticks to more concrete positions when inter-institutional conflict threatens its policy goals but employs vagueness to maintain a broad appeal and avoid public backlash given public opinion divisions and contentiousness. These results add to our knowledge of legal text formulation, EU law making, and EP representation.

Do wars follow predictable patterns, or are they chaotic and inherently unpredictable? We explore this question by quantifying temporal redundancy—how often similar patterns recur—in conflict-related fatalities. Using entropy-based analysis—a measure of randomness—and Dynamic Time Warping, we compare conflict dynamics with other complex systems such as seismology and finance. Our findings reveal that the dynamics of fatalities within conflicts exhibit high levels of entropy: they are largely unpredictable. Conflicts generally display higher entropy than climate or earthquake data, which highlights the challenges of predictive modeling. Moreover, conflict patterns are not region- or decade-specific, which suggests a degree of universality. Our framework distinguishes stable from chaotic phases in conflict dynamics and clarifies where forecasting is feasible.