It is often argued that when legislators have personal vote-seeking incentives, parties are less unified because legislators need to build bonds of accountability with their voters. I argue that these effects depend on a legislator’s ability to cultivate a personal vote. When parties control access to the ballot and the resources candidates need to cultivate personal votes, they can condition a legislator’s access to these resources on loyalty to the party’s agenda. I test this theory by conducting a difference-in-differences analysis that leverages the staggered implementation of the 2014 Mexican Electoral Reform. This reform introduced the possibility of consecutive reelection for state legislators, increasing their incentives to cultivate personal votes. I study unity in position-taking and voting behaviour of Mexican state legislators from 2012 to 2018. To analyze position-taking, I apply correspondence analysis to a new dataset of over half a million legislative speeches in twenty states. To study voting, I analyze over 14,500 roll-call votes in fourteen states during the same period. Results show that reelection incentives increased intra-party unity, which has broad implications for countries introducing electoral reforms aiming to personalize politics.

A common challenge in studying Italian parliamentary discourse is the lack of accessible, machine-readable, and systematized parliamentary data. To address this, this article introduces the ItaParlCorpus dataset, a new, annotated, machine-readable collection of Italian parliamentary plenary speeches for the Camera dei Deputati, the lower house of Parliament, spanning from 1948 to 2022. This dataset encompasses 470 million words and 2.4 million speeches delivered by 5830 unique speakers representing 77 different political parties. The files are designed for easy processing and analysis using widely-used programming languages, and they include metadata such as speaker identification and party affiliation. This opens up opportunities for in-depth analyses on a variety of topics related to parliamentary behavior, elite rhetoric, and the salience of political themes, exploring how these vary across party families and over time.

Previous accounts have suggested a potential divergence between Xi Jinping and Li Keqiang in their approaches to economic governance. This study examines the policy orientations of the two leaders concerning state–market relations, providing empirical evidence for the recent manifestation of what insiders have termed the “dispute between north and south houses” (nanbeiyuan zhi zheng) and its economic implications. By applying semi-supervised machine learning methods to textual data, this study demonstrates that Li favoured market-oriented policies, whereas Xi displayed a pronounced preference for state-centric strategies. The findings notably indicate an initial divergence in policy orientation, which was followed by a considerable convergence during Xi's second term. Our analysis further reveals that Li's market-oriented rhetoric was particularly prominent during “Mass innovation week,” indicating a campaign-style policy mobilization. Moreover, the analysis identifies that the discursive differences between the two leaders are associated with a decline in firm-level investment, suggesting that disparities in policy orientation may engender political uncertainty. This study contributes to the extant literature on the impact of leadership dynamics on economic policy, the implications of mixed signals from the central leadership and the phenomenon of campaign-style mobilization in China.

We apply moral foundations theory (MFT) to explore how the public conceptualizes the first eight months of the conflict between Ukraine and the Russian Federation (Russia). Our analysis includes over 1.1 million English tweets related to the conflict over the first 36 weeks. We used linguistic inquiry word count (LIWC) and a moral foundations dictionary to identify tweets’ moral components (care, fairness, loyalty, authority, and sanctity) from the United States, pre- and post-Cold War NATO countries, Ukraine, and Russia. Following an initial spike at the beginning of the conflict, tweet volume declined and stabilized by week 10. The level of moral content varied significantly across the five regions and the five moral components. Tweets from the different regions included significantly different moral foundations to conceptualize the conflict. Across all regions, tweets were dominated by loyalty content, while fairness content was infrequent. Moral content over time was relatively stable, and variations were linked to reported conflict events.

Large language models are a powerful tool for conducting text analysis in political science, but using them to annotate text has several drawbacks, including high cost, limited reproducibility, and poor explainability. Traditional supervised text classifiers are fast and reproducible, but require expensive hand annotation, which is especially difficult for rare classes. This article proposes using LLMs to generate synthetic training data for training smaller, traditional supervised text models. Synthetic data can augment limited hand annotated data or be used on its own to train a classifier with good performance and greatly reduced cost. I provide a conceptual overview of text generation, guidance on when researchers should prefer different techniques for generating synthetic text, a discussion of ethics, a simple technique for improving the quality of synthetic text, and an illustration of its limitations. I demonstrate the usefulness of synthetic training through three validations: synthetic news articles describing police responses to communal violence in India for training an event detection system, a multilingual corpus of synthetic populist manifesto statements for training a sentence-level populism classifier, and generating synthetic tweets describing the fighting in Ukraine to improve a named entity system.

Content analysis is a valuable tool for analysing policy discourse, but annotation by humans is costly and time consuming. ChatGPT is a potentially valuable tool to partially automate content analysis for policy debates, largely replacing human annotators. We evaluate ChatGPT’s ability to classify documents using pre-defined argument descriptions, comparing its performance with human annotators for two policy debates: the Universal Basic Income debate on Dutch Twitter (2014–2016) and the pension reforms debate in German newspapers (1993–2001). We use the API (GPT-4 Turbo) and user interface version (GPT-4) and evaluate multiple performance metrics (accuracy, precision and recall). ChatGPT is highly reliable and accurate in classifying pre-defined arguments across datasets. However, precision and recall are much lower, and vary strongly between arguments. These results hold for both datasets, despite differences in language and media type. Moreover, the cut-off method proposed in this paper may aid researchers in navigating the trade-off between detection and noise. Overall, we do not (yet) recommend a blind application of ChatGPT to classify arguments in policy debates. Those interested in adopting this tool should manually validate bot classifications before using them in further analyses. At least for now, human annotators are here to stay.

Populist radical right (PRR) parties' attacks against prevailing historical interpretations have received much public attention because they question the foundations of countries' political orders. Yet, how prominent are such attacks and what characterizes their sentiment and content? This article proposes an integrated mixed-methods approach to investigate the prominence, sentiment, and interpretations of history in PRR politicians' parliamentary speeches. Studying the case of Germany, we conducted a quantitative analysis of national parliamentary speeches (2017–2021), combined with a qualitative analysis of all speeches made by Alternative for Germany (AfD) in 2017–2018. The AfD does not use historical markers more prominently but is distinctly less negative when speaking about history compared to its general political language. The collocation and qualitative analyses reveal the nuanced ways in which the AfD affirms and disavows various mnemonic traditions, underlining the PRR's complex engagement with established norms.

An extensive theoretical and practitioner literature addresses the drivers and consequences of transformation of violent rebel actors during conflicts. However, measurement challenges constrain large-N empirical study of the effects and consequences of such transformations. This Research Note introduces a strategy to identify periods of transformation and change in the operation of non-state armed militant groups via computational text analysis of trends in reporting on activities. It presents the measurement approach and demonstrates scalability to a corpus of more than 200 militant groups operating from 1989 to 2020. The study concludes by extending a recent analysis of the impacts of uncertainty on conflict termination. An online Appendix demonstrates the advantages and drawbacks of the measurement through a series of case studies.

Are centralized leaders of religious organizations responsive to their followers' political preferences over time even when formal accountability mechanisms, such as elections, are weak or absent? I argue that such leaders have incentives to be responsive because they rely on dedicated members for legitimacy and support. I test this theory by examining the Catholic Church and its centralized leader, the Pope. First, I analyze over 10,000 papal statements to confirm that the papacy is responsive to Catholics' overall political concerns. Second, I conduct survey experiments in Brazil and Mexico to investigate how Catholics react to responsiveness. Catholics increase their organizational trust and participation when they receive papal messages that reflect their concerns, conditional on their existing commitment to the Church and their agreement with the Church on political issues. The evidence suggests that in centralized religious organizations, the leader reaffirms members' political interests because followers support religious organizations that are politically responsive.

While presidents frequently create new policies through unilateral power, empirical scholarship generally focuses on executive orders and overlooks other categories of directives. We introduce data on more than 50,000 unilateral directives issued between 1877 and 2020 and use machine learning techniques to characterize their substantive importance and issue areas. Our measures reveal significant increases in unilateral activity over time, driven largely by increases in foreign affairs and through the substitution of memoranda for executive orders. We use our measures to formally evaluate the historical development of the unilateral presidency and reassess theoretical claims about public opinion and unilateral power. Our research provides new evidence about variation in the use of presidential authority and opens new avenues for empirical inquiry.

This paper examines the strategic use of public news media – specifically television (TV) – as an instrument of political influence, focusing on Italy's 2011 financial crisis under Berlusconi's premiership. Using an original large corpus of over 20,000 hours of televised news transcripts and a quasi-experimental design, we investigate how political influence altered media coverage and, subsequently, public opinion and electoral outcomes. Our difference-in-differences analysis, complemented by unsupervised text scaling of news content, reveals a significant shift from “hard” political news to “soft” news on public TV during Berlusconi's tenure. Findings suggest a deliberate reduction in hard news coverage by an average of 107 seconds daily, which significantly increased voter support for Berlusconi's party. In the conclusions, we discuss the broader implications of our findings for media independence in Western democracies amid the emergence of artificial intelligence-generated news contents and the prevalence of algorithmically tailored news feeds.

Female attorneys at the U.S. Supreme Court are less successful than male attorneys under some conditions because of gender norms, implicit expectations about how men and women should act. While previous work has found that women are more successful when they use more emotional language at oral arguments, gender norms are context sensitive. The COVID-19 pandemic prompted perhaps the most radical contextual shift in Supreme Court history: freewheeling in-person arguments were replaced with turn-based teleconference arguments. This change altered judicial decision-making and, I argue, justices’ assessments of attorneys’ gender performance. Using quantitative textual analysis of oral arguments, I demonstrate that justices implicitly evaluate gender performance with different metrics in each modality. Gender-normative levels of emotional language predict success in both formats. Function words, however, only predict success in teleconference arguments. Given gender’s salience at the Supreme Court and in broader society, my findings prompt questions about the extent to which women can substantively impact case law.

We investigate the gender gap in issue attention among members of parliament (MPs) by applying automated text analytic techniques to a novel data set on Italian parliamentary speeches over a remarkably long period (1948–2020). We detect a gendered specialization across issues that tends to disappear as women’s shares in parliamentary groups increase. We then investigate whether women’s access to previously male-owned issues brings with it a different agenda, operationalized as a different vocabulary. We detect a U-shaped pattern: language gender specificity is high when female MPs are tokens in parliamentary groups with a large preponderance of men; it decreases when their shares start increasing and grows again when they constitute a considerable minority. We argue that this pattern is consistent with the theory of tokenism, and it is produced by the interlinkage of commitment to shared norms and the distribution of “activation thresholds” among female MPs.

Extensive research in Western societies has demonstrated that media reports of protests have succumbed to selection and description biases, but such tendencies have not yet been tested in the Chinese context. This article investigates the Chinese government and news media's selection and description bias in domestic protest events reporting. Using a large protest event data set from Weibo (CASM-China), we found that government accounts on Weibo covered only 0.4 per cent of protests while news media accounts covered 6.3 per cent of them. In selecting events for coverage, the news media accounts tacitly struck a balance between newsworthiness and political sensitivity; this led them to gravitate towards protests by underprivileged social groups and shy away from protests targeting the government. Government accounts on Weibo, on the other hand, eschewed reporting on violent protests and those organized by the urban middle class and veterans. In reporting selected protest events, both government and news media accounts tended to depoliticize protest events and to frame them in a more positive tone. This description bias was more pronounced for the government than the news media accounts. The government coverage of protest events also had a more thematic (as opposed to episodic) orientation than the news media.

It is well known that politicians speak differently when campaigning. The shadow of elections may affect candidates' change in tone during campaigns. However, to date, we lack a systematic study of the changes in communication patterns between campaign and non-campaign periods. In this study, we examine the sentiment expressed in 4.3 million tweets posted by members of national parliaments in the EU27 from 2018 to 2020. Our results show that (1) the opposition, even populists and Eurosceptics, send more positive messages during campaigns, (2) parties trailing in the polls communicate more negatively, and (3) that the changes are similar in national and European elections. These findings show the need to look beyond campaign times to understand parties' appeals and highlight the promises of social media data to move beyond traditional analyses of manifestos and speeches.

Economists have long been interested in the effect of business sentiment on economic activity. Using text analysis, I construct a new company-level indicator of sentiment based on the net balance of positive and negative words in Australian company disclosures. Company-level investment is very sensitive to changes in this corporate sentiment indicator, even controlling for fundamentals, such as Tobin’s Q, as well as controlling for measures of company-level uncertainty.

The high sensitivity of investment to sentiment could be due to several mechanisms. It could be because of animal spirits among managers or because of sentiment proxies for private information held by managers about company prospects. Overall, I find mixed evidence of the underlying causal mechanism. The effect of sentiment on investment is relatively persistent, which is consistent with managers having private information about company fundamentals. But the sensitivity of investment to sentiment is not any stronger at opaque companies in which managers are likely to be better informed than investors. Further, investment is sensitive to sentiment even when investors have an information advantage over managers by lagging the sentiment indicator by a year. Overall, the sensitivity of investment to sentiment appears to reflect both animal spirits and fundamentals.

Corporate investment has been weak since the global financial crisis (GFC) and demand-side factors, such as lower sales growth, explain more than half of this weakness. Low sentiment and heightened uncertainty weighed on investment during the GFC but have been less important factors since then.

Research of judges and courts traditionally centers on judgments, treating each judgment as a unit of observation. However, judgments often address multiple distinct and more or less unrelated issues. Studying judicial behavior on a judgment level therefore loses potentially important details and risks drawing false conclusions from the data. We present a method to assist researchers with splitting judgments by issues using a supervised machine learning classifier. Applying our approach to splitting judgments by the Court of Justice of the European Union into issues, we show that this approach is practically feasible and provides benefits for text-based analysis of judicial behavior.