This study investigates whether partisans dynamically change their news consumption in response to negative coverage about their own party or the opposing party—a phenomenon labeled partisan temporal selective news avoidance. The analysis uses large-scale online behavior to capture real-world browsing adjustments to changing sentiment in the news environment.
This paper asks whether US economic sanctions during the Trump presidency reduced support for Iran’s government or instead produced a "rally-around-the-flag" effect among Iranian influencers and their audiences. Using a large-scale supervised machine-learning analysis of nearly two million tweets from more than 1,000 Iranian influencers, the authors compare reactions to comprehensive versus targeted sanctions and examine variation by political faction and by which entity was targeted.
What explains frequent exchange-rate interventions in East Asia? This study asks whether organized export interests prompt policymakers to devalue currencies and presents evidence from Japan and Korea using news-based measures of exporters’ public pressure. The authors build a proxy for exporter lobbying from daily reports in leading financial newspapers and apply machine-learning scaling to trace how industry demands evolve before interventions. Their contribution is a novel measurement strategy linking publicized industry demands to episodes of currency depreciation.
A study of the Proud Boys' Telegram communications and the group's real-world actions over a 31-month period, investigating how different kinds of online speech relate to subsequent offline behavior.
This study asks how political parties divide attention across the past, present and future in campaign communication and whether incumbency shapes emotive rhetoric across these temporal dimensions. Using supervised machine learning to label temporal focus, the author analyzes 621 party manifestos from nine countries spanning 1949–2017 and describes the share of manifesto text devoted to each temporal category. The contribution is a systematic, cross-national description of parties' temporal rhetorical emphasis and how statements about past and present relate to differences in sentiment between incumbents and opposition.
Many scholars classify documents into categories, but social scientists often need accurate estimates of the proportion of documents in each category. Parametric “classify-and-count” approaches can be highly model-dependent and may produce greater bias in proportion estimates even when the percent of documents correctly classified rises. Direct, nonparametric estimation of proportions avoids some of those model-dependence problems but can fail when language meaning shifts between training and test sets or when categories use very similar language.
Convolutional neural networks (CNNs) are presented as a practical tool for classifying visual information in social science research. The method offers a way to speed up the tedious task of labeling images and extracting structured information from visual documents, making new data sources usable for analysis and policy work.
A constrained priority mechanism is introduced that blends outcome-based matching (from machine learning predictions) with traditional preference-based allocation. The design lets a planner enforce a minimum average outcome while still assigning agents according to their ranked preferences.
This research investigates how policymakers interpret signals during high-stakes diplomacy, focusing on the 1958-63 standoff. Contrary to conventional wisdom that emphasizes public and private communications, findings demonstrate material actions exert a stronger influence.
Politics and political conflict are often expressed in written and spoken words, but the massive costs of analyzing even moderately sized text collections have limited their use in political science. Automated text analysis substantially reduces those costs and has already delivered on part of that promise, making large-scale study of political texts more feasible.
The surge of reports from monitors like Amnesty International, Human Rights Watch, and the US State Department offers unprecedented detail on repression and rights protections—but most automatic approaches reduce texts to simple word counts that ignore syntax and word order, limiting what scholars and policy-makers can learn.
This study investigates who starts political disputes by examining a unique dataset of ethnic conflicts across Africa. Using advanced statistical modeling techniques applied to original data collection, the research reveals an unexpected pattern: democratic transitions actually increase conflict initiation rates in some contexts.