Stefan Müller
LightNews LightNews
banner
stefanmueller.bsky.social
Stefan Müller
@stefanmueller.bsky.social
3.9K followers 1.5K following 240 posts
Associate Professor @ucddublin.bsky.social • party competition, public opinion, political communication, computational social science • Maintainer @irishpollingind.bsky.social • Executive Committee Member @yai.ie
muellerstefan.net
Posts Replies Media Videos
stefanmueller.bsky.social
Really important and insightful analysis of the voters who spoiled their ballots in the presidential election.

@kevcunningham.bsky.social also classifies open-ended survey responses to better understand the underlying reasons and motivations.

➡️ open.substack.com/pub/kevcunni...
Respondents were asked ‘Why’ they voted the way they did in an open-ended text box. These are the responses classified into groups. Open-ended survey responses – proportional sample of statements % of each voter type that spoiled their ballot as opposed to those who voted at the 2025 Irish Presidential election. % who spoiled their vote broken down by age, gender and self-reported social class.
October 29, 2025 at 1:39 PM
stefanmueller.bsky.social
Reasons for vote choice according to the Ireland Thinks polling day poll ⤵️
#Aras25
analysis.irelandthinks.ie/polling-day-...
@kevcunningham.bsky.social
Which of the following was most important in making your choice? (Select up to 2) Connolly voters relied on Integrity & Honesty, Competence and Communication skills. Humphreys voters more likely to focus on how the candidate represented Ireland internationally.
October 25, 2025 at 11:10 AM
stefanmueller.bsky.social
📊 @kevcunningham.bsky.social is publishing findings from the Ireland Thinks polling day poll at analysis.irelandthinks.ie
Question: Thinking about your number 1, first- preference vote, was that choice mainly FOR your chosen candidate, OR mainly AGAINST another candidate, or both about equally? Question: If elected, do you feel like Catherine Connolly would be similar to Michael D. Higgins as president?
October 25, 2025 at 10:37 AM
stefanmueller.bsky.social
There goes the paper on support for Jim Gavin among those interested in Gaelic games in Dublin versus rival counties.

Liam Kneafsey and I successfully applied for including these questions in the Irish Presidential Election Study, and we completed the pre-analysis plan and code last week. 😬😢
Pre-Analysis Plan: Rivalries and Celebrity Candidates: How Sport Shapes Voting Behaviour in Ireland Authors: Liam Kneafsey and Stefan Müller
October 6, 2025 at 10:28 AM
stefanmueller.bsky.social
New paper in Computational Communication Research with James Cross, @derekgreene.bsky.social & Martijn Schoonvelde: “Mapping Digital Campaign Strategies: How Political Candidates Use Social Media to Communicate Constituency Connection and Policy Stance”

PDF (open access): doi.org/10.5117/CCR2...
Abstract: Social media have become a crucial tool for candidates seeking election, allowing them to build a public profile by posting curated content to appeal to potential voters. Focusing on the 2020 Irish General Election, this study investigates how candidates used Twitter to signal their campaign efforts and policy positions, and how their communicative priorities varied based on their gender, competitiveness, and political experience. To do so, we first demonstrate that a transformer-based machine-learning approach based on sentence embeddings can successfully identify social media posts that contain policy and electioneering content. Our findings show that experienced candidates are more likely to emphasise policy-related content than less experienced ones. This pattern also holds for electioneering content when we account for previous engagement with such posts. Contrary to our pre-registered expectations, we find no meaningful differences in the emphasis on electioneering or policy content based on candidates' gender or electoral competitiveness. Overall, our results demonstrate how candidates strategically use social media to shape their public personas during election campaigns in Ireland's candidate-centred electoral system, with multi-member constituencies and strict campaign spending limits.
September 17, 2025 at 10:41 AM
stefanmueller.bsky.social
Are you a @ucdpolitics.bsky.social graduate? Join our discussion event on the upcoming presidential election. 

🎙️ Panelists: @kevinrafter.bsky.social, Art O'Leary & Sheila Naughton
🗓️ 30 Sept, 18:30
📍 Museum of Literature Ireland

Register here: docs.google.com/forms/d/e/1F...
@ucddublin.bsky.social
Ireland's Presidency: More Than a Figurehead? UCD School of Politics and International Relations SPIRe alumni in conversation with Art O'Leary, Executive of An Coimisiún Toghcháin/The Electoral Commission Kevin Rafter, Full Professor of Political Communication at DCU Moderator: Sheila Naughton, RTÉ RSVP and join us Tuesday, September 30 6:30pm, Museum of Literature Ireland
September 16, 2025 at 10:15 AM
stefanmueller.bsky.social
Some of the findings:
– Quality of training data is more important than quantity
– Separate binary classifiers performed best
– Grant peer review is much more positive than negative

We are now working on substantive applications (see thread below).

Classifiers: huggingface.co/snsf-data
Pipeline for identifying, classifying, and validating textual characteristics in grant peer review reports Keyness analysis revealing predictive terms for each category in the full sample of project funding review The predicted prevalence of each category on the level of reviews. Vertical dashed lines show the average prevalence across the 47,522 review reports. F1 scores (macro-average) depending on the size of the training set
September 3, 2025 at 9:11 AM
stefanmueller.bsky.social
Our paper “A Supervised Machine Learning Approach for Assessing Grant Peer Review Reports” has been accepted at Quantitative Science Studies.

We present a workflow and fine-tuned transformer models to analyse grant peer review reports at scale.

Paper (open access): doi.org/10.1162/qss....
Abstract: Peer review is essential to the research lifecycle, yet the contents of grant peer review reports remain underexplored. Our study addresses this gap by developing a pipeline to systematically analyze these reports using Natural Language Processing and Machine Learning. We define twelve categories relevant to funding agencies, create an annotation codebook, fine-tune and validate transformer models, and apply these classifiers to a novel text corpus consisting of 1.6 million sentences from 47,522 grant peer review reports submitted to the Swiss National Science Foundation. This work has critical implications for the academic community. It provides novel insights into the content of grant peer review reports and openly available tools to enhance transparency, fairness, and consistency in grant evaluation. Our findings also highlight differences between journal and grant peer reviews, while the developed framework enables funding agencies and researchers to refine practices, fostering a more trustworthy and efficient evaluation process.
September 3, 2025 at 9:11 AM
stefanmueller.bsky.social
Pleased to announce the Connected_Politics Lab Autumn Seminars:

– @profdansmith.bsky.social & Nam Pham
– Eoghan Cunningham @ucdpolitics.bsky.social
– Elissa Berwick @mcgill.ca
– @jbslapin.bsky.social
– @sabrinabarias.bsky.social
– @dpzollinger.bsky.social

All welcome!
www.ucd.ie/connected_po...
AUTUMN TERM 2025 Gender, Dynastic Ties, and Representational Style in Legislatures Speaker: Nam Pham and Daniel M. Smith (University of Pennsylvania) Wednesday, September 17, 14:00–14:45 (Irish time) Measuring and Mitigating Bias in Parliamentary Discourse Summarisation Speaker: Eoghan Cunningham (University College Dublin) Wednesday, October 1, 14:00–14:45 (Irish time) The Trouble with Cohorts: Reconsidering Generational Differences in Democratic Deconsolidation Speaker: Elissa Berwick (McGill University) Wednesday, October 22, 14:00–14:45 (Irish time) How Descriptive Over- and Under-Representation Impacts Citizens Evaluations of Decision-making across Policy Domains Speaker: Jonathan Slapin (University of Zurich) Wednesday, November 5, 14:00–14:45 (Irish time) Everything Old is New Again: Textual Recycling in UN Resolutions Speaker: Sabrina Arias (Lehigh University) Wednesday, November 12, 14:00–14:45 (Irish time) ‘Undeserved winners’ or ‘more deserving losers’ ? Two types of societal discontent and their relation to radical right and left voting Speaker: Delia Zollinger (University of Zurich) Wednesday, November 26, 14:00–14:45 (Irish time)
August 25, 2025 at 11:21 AM
stefanmueller.bsky.social
Andy Eggers (@aeggers.bsky.social) and I are co-chairing the Political Methodology section for EPSS 2026 in Belfast.

Please submit a paper, full panel proposals, or suggestions for alternative formats (such as roundtables and lightning rounds).

➡️ epssnet.org/belfast-2026...
@epssnet.bsky.social
EPSS panel formats 2026 Standard Panels These panels focus on the presentation of full research papers. Each panel can include up to five papers and must have one chair and one or two discussants. All panelists must confirm their participation in advance. As a variant chairs might coordinate with their panels to try a flipped approach in which the presenter does not present at all; instead discussants (possibly as many as one per paper) are tasked with both presenting the work and critiquing it Lightning Rounds Fast-paced and engaging, lightning rounds follow a PechaKucha-style format (e.g. 15 slides, x 20 seconds each). Each of 8 presenters gets 5 minutes to share research, followed by 5 minutes of feedback or Q&A. A chair leads with support from 2 discussants. Ideal for sharpening communication skills and getting rapid feedback on emerging work. Author meets critics A panel discussion centered on a major scholarly work—usually a book or article. Panelists will explain the work’s key contributions, offer critiques, and engage in thoughtful debate. This is a great format for deeper engagement with influential scholarship. Topical round tables A panel discussion centered on a topical issue related to the profession or salient political events. Consider involving politicians and practitioners.
August 5, 2025 at 1:59 PM
stefanmueller.bsky.social
Excellent initiative: The Irish Presidential Election Study (2025) seeks proposals for questions on Irish elections & democracy.

– 3 waves: campaign, election week, post-election
– Submit proposal incl. theory and evidence to Gail McElroy
– Deadline: 12 Sept 2025

@psaireland.bsky.social
neds.ie
Call for Questions: Irish Presidential Election Study The NEDS Management Board invites submissions on questions for inclusion in the Irish Presidential Election Study, which is expected to take place in October 2025. There will be 3 waves- one at the beginning of the campaign, one during the week of the election and one immediately after the election. Researchers are encouraged to submit proposals for survey questions. Questions should be directly related to Irish elections and democracy. Proposals may include single items or short question blocks. Please note that survey space is highly limited, so being concise is essential. As such, shorter single item questions will be viewed more favourably. Researchers proposing a battery of questions should indicate the relative priority of each item. The maximum length for a battery submission is 5-minutes. Researchers are also asked to indicate which wave they would like their question(s) included in. Submission Guidelines Each proposal should be no more than three pages and must include: · Abstract: A concise summary of the proposed question(s). · Theoretical Justification: A clear explanation of the underlying theory and how it connects with existing research in the field. · Question Wording: Exact phrasing of the proposed question(s). · Supporting Evidence: Details of any pre-tests conducted and/or international use of similar questions. · Relevance and Comparability: A statement on the broader applicability of the questions across contexts and their comparability with other datasets. Deadline for questions is 3pm Friday September 12th, 2025. For any queries or to submit your proposal, please contact: Prof Gail McElroy
July 22, 2025 at 8:18 AM
stefanmueller.bsky.social
The uncorrected proofs have arrived!

Gallagher, O‘Malley, and @theresareidy.bsky.social: “How Ireland Voted 2024 – The New Normal?”
📘 link.springer.com/book/9783031...
Chapter 5 Gail McElroy and Stefan Müller What the 2024 Manifestos Reveal: Issue Salience and Left–Right Positions Historically dominated by two centre-right parties, Ireland’s political landscape has long puzzled international commentators and academics. Three recurring questions arise: Why has the left traditionally been so weak? Why is there no far-right party represented in Dáil Éireann? And perhaps the biggest conundrum of all, what truly sets Fianna Fáil apart from Fine Gael? These questions have contributed to Ireland’s reputation as an outlier among its western and northern European peers, with a party system often described as sui generis. Recently, however, there have been signs of a more classic left–right dimension emerging, and the 2020 election marked a significant turning point when a party of the left won a plurality of the votes. In this chapter, we examine the evolving nature of the Irish policy space by analysing the 2024 election manifestos of all the main political parties. Book cover: How Ireland Voted 2024
July 19, 2025 at 6:20 AM
stefanmueller.bsky.social
Paper (open access): doi.org/10.1017/psrm...

Replication materials and digitised manifestos: doi.org/10.7910/DVN/...

7/7
We thank Shaun Bowler, Frederik Hjorth, Sarah King, Joseph Lacey, Yen-Chieh Liao, Robin Rauner, Chiara Valsangiacomo, the editor and anonymous reviewers, and participants at the Annual Conference of the European Political Science Association in Prague, the Politics Research Colloquium at the University of Oxford, the workshop on Election Campaigns and Democratic Norms in Dublin, and the SPIRe Seminar Series at University College Dublin for their comments and improvement suggestions. We are grateful to Yutaka Shinada for sharing digitized candidate manifestos. We would like to thank Daisuke Hakiai, Keiko Higuchi, Momoho Kikukawa, Kana Miyazaki, Tsubasa Nishimura, and Ken Yoshioka for outstanding research assistance.
July 15, 2025 at 7:43 AM
stefanmueller.bsky.social
Second key finding: Congruence is higher for distributive policy areas and for issues deemed less important, based on expert survey assessments. These results suggest that campaigns function as meaningful signals of future legislative priorities.

6/7
Predicted probability of obtaining a post (Combined Measure) conditional on the interaction effect between portfolio importance and varying levels of manifesto salience. Coefficient estimates and 95 percent confidence intervals of Manifesto Salience based on separate logistic regression models for each policy area.
July 15, 2025 at 7:43 AM
stefanmueller.bsky.social
First key finding: A stronger focus on a policy issue during the campaign significantly increases the probability of securing a legislative leadership post in that area.

This pattern is robust across modelling choices and measures of issue emphasis, and it holds for every election.

5/7
Predicted probabilities of obtaining legislative leadership posts conditional on the salience of the same policy area in candidate manifestos.
July 15, 2025 at 7:43 AM
stefanmueller.bsky.social
In the next step, we match the emphasis on policy issues with novel data on all legislative leadership posts from 2003 to 2017. The data collected by ‪@profdansmith.bsky.social & Steven Reed allow us to consider many relevant characteristics of MPs in our analyses.

4/7
Overview of classification and aggregation procedure
July 15, 2025 at 7:43 AM
stefanmueller.bsky.social
After fine-tuning a transformer machine learning model (Japanese BERT), we classify policy areas within candidate manifestos published between 2003 and 2014. Out-of-sample predictions closely match human coding, and additional tests support their face validity.

3/7
Correspondence between the frequency of policy areas in the test set based on the BERT predictions and human coding of the same set of statements.
July 15, 2025 at 7:43 AM
stefanmueller.bsky.social
Do candidates’ policy priorities during election campaigns predict their legislative activities? Our study focuses on Japanese politicians, combining new data on legislative posts with over 46,000 statements from 1,270 candidate manifestos across five elections.

2/7
Three typical candidate manifestos
July 15, 2025 at 7:43 AM
stefanmueller.bsky.social
‘Campaign Communication and Legislative Leadership’ (with Naofumi Fujimura) appeared in the latest issue of Political Science Research and Methods.

Paper (open access): doi.org/10.1017/psrm...
@psrm.bsky.social

Thread below 🧵 1/7
Abstract Do policy priorities that candidates emphasize during election campaigns predict their subsequent legislative activities? We study this question by assembling novel data on legislative leadership posts held by Japanese politicians and using a fine-tuned transformer-based machine learning model to classify policy areas in over 46,900 statements from 1270 candidate manifestos across five elections. We find that a higher emphasis on a policy issue increases the probability of securing a legislative post in the same area. This relationship remains consistent across multiple elections and persists even when accounting for candidates' previous legislative leadership roles. We also discover greater congruence in distributive policy areas. Our findings indicate that campaigns provide meaningful signals of policy priorities.
July 15, 2025 at 7:43 AM
stefanmueller.bsky.social
Received a complimentary copy of “Foundations of European Politics” after commenting on new chapters. Such a good textbook: clear theory, empirical evidence, and free interactive data exercises.

@catherinedevries.bsky.social @sarahobolt.bsky.social @sopro.bsky.social @jbslapin.bsky.social
Book “Foundations of European Politics” Vote shares of populist parties across Europe Frequency of majority, minority, and caretaker governments in Europe Table of contents
July 3, 2025 at 11:48 AM
stefanmueller.bsky.social
Results from regression models correspond to the descriptive findings. Female reviewers focus more on all evaluation criteria and are more positive. Female applicants receive slightly more positive and less negative reviews than male applicants. Differences by research domain are substantive.

10/12
List of variables included in the regression analysis
July 2, 2025 at 8:08 AM
stefanmueller.bsky.social
The differences in positive/negative statements across disciplines are fascinating. Reviews in MINT disciplines are much more positive and much less negative than reviews in LS and SSH.

8/12
Average prevalence of positive (A) and negative (B) statements (% of review sentences) by discipline. Horizontal bars show 95% confidence intervals. Separate horizontal axis limits to allow for comparisons within each content category. Overall, reviews tend to be much more positive than negative.
July 2, 2025 at 8:08 AM
stefanmueller.bsky.social
Next, we turn to the review content. We classify over 1.3 million sentences across the four SNSF evaluation criteria as well as sentiment. The table below shows “typical” words for each content category. We then aggregate the prevalence of each category to the level of review reports.

6/12
Frequent predictive terms for each content category. The results are based on keyness analyses using χ² tests for each word or multi-word expression, comparing frequencies in sentences where a content characteristic was present (target group) with those where it was absent (reference group).
July 2, 2025 at 8:08 AM
stefanmueller.bsky.social
We find that female reviewers write considerably longer reviews than male reviewers across all research domains.

Female reviewers also receive somewhat longer reports than male reviewers, especially in SSH.

5/12
Average review length by the gender of reviewers and gender of applicants. Dots show mean length (in words), horizontal bars depict 95% confidence intervals. Average review length for reports submitted by male and female reviewers (A) and male and female applicants (B). Dots show mean length (in words), horizontal bars depict 95% confidence intervals.
July 2, 2025 at 8:08 AM
stefanmueller.bsky.social
Let’s start with review length: reviews in the Social Sciences and Humanities (SSH) are considerably longer than reviews in the Life Sciences (LS) and MINT (Mathematics, Informatics, Natural Sciences, Technology). The plot below underscores these substantive differences across disciplines.

4/12
Average review length conditionally on the research domain. Dots show the mean length (in words), horizontal bars the 95% confidence intervals
July 2, 2025 at 8:08 AM