I am a Postdoctoral Researcher in the Department of Political Science at the University of Zurich. I received my PhD in Political Science from Trinity College Dublin. I am a core contributor to the quanteda R package, Training Advisor of the Quanteda Initiative, and author of extensive tutorials on quantitative text analysis. I am also affiliated with the Digital Democracy Lab at the University of Zurich.
My research focuses on the interactions between parties, voters, and the media. More precisely, I reassess the concept of election pledges, analyse the circumstances under which political parties decide to make promises, claim credit or attribute blame, and examine how the media report on election promises. Additionally, I am involved in projects on coalition prediction, public opinion, legislative behaviour, the incumbency advantage, and political compromise.
Methodologically, I develop and validate tools for the efficient and reliable combination of human coding and machine learning. Moreover, with an interdisciplinary team of researchers at the Swiss National Science Foundation, I leverage supervised classification to measure the quality and transparency of peer review.
My work has been published in Political Science Research and Methods, Electoral Studies, the Journal of Elections, Public Opinion and Parties, the International Journal of Performance Analysis in Sport, and the Journal of Open Source Software.
Stefan Müller and Tom Louwerse. Forthcoming. “The Electoral Cycle Effect in Parliamentary Democracies.” Political Science Research and Methods (online first).
Does government party support decline in a monotonic fashion throughout the legislative cycle or do we observe a u-shaped ‘electoral cycle effect’? Moving beyond the study of midterm election results, this is the first study to assess the cyclical pulse of government party support in parliamentary democracies based on over 25,000 voting intention polls from 171 cycles in 22 countries. On average, government parties lose support during the first half of the electoral cycle, but at most partially recover from their initial losses. Under single-party government and when prime ministers control cabinet dissolution, support tends to follow the previously assumed u-shaped pattern more strongly. Finally, we find that government parties hardly recover from early losses since the 2000s.
Stefan Müller and Michael Jankowski. 2019. “Do Voters Really Prefer More Choice? Determinants of Support for Personalised Electoral Systems.” Journal of Elections, Public Opinion and Parties 29(2): 262–281.
Which voters prefer having more choice between parties and candidates in an election? To provide an answer to this question, we analyse the case of a radical change from a closed-list PR system to a highly complex open-list PR system with cumulative voting in the German states of Bremen and Hamburg. We argue that the approval of a personalised electoral system is structured in similar ways as support for direct democracy. Using representative surveys conducted prior to all four state elections under cumulative voting in 2011 and 2015, we analyse which individual factors determine the approval, disapproval or indifference towards the new electoral law. The results indicate that younger voters as well as supporters of left parties are much more likely to support a personalised electoral system. In contrast to previous studies, political interest only has an impact on the indifference towards the electoral system. More generally, our results show that a large proportion of voters does not appreciate personalised preferential electoral systems which seems to be a result of the complexity and magnitude of choice between parties and candidates.
Kenneth Benoit, Kohei Watanabe, Haiyan Wang, Paul Nulty, Adam Obeng, Stefan Müller, and Akitaka Matsuo. 2018. “quanteda: An R Package for the Quantitative Analysis of Textual Data.” Journal of Open Source Software 3(30): 774.
quanteda is an R package providing a comprehensive workflow and toolkit for natural language processing tasks such as corpus management, tokenization, analysis, and visualization. It has extensive functions for applying dictionary analysis, exploring texts using keywords-in-context, computing document and feature similarities, and discovering multi-word expressions through collocation scoring. Based entirely on sparse operations, it provides highly efficient methods for compiling document-feature matrices and for manipulating these or using them in further quantitative analysis. Using C++ and multi-threading extensively, quanteda is also considerably faster and more efficient than other R and Python packages in processing large textual data.
Shaun Bowler, Gail McElroy, and Stefan Müller. 2018. “Voter Preferences and Party Loyalty under Cumulative Voting: Political Behaviour after Electoral Reform in Bremen and Hamburg.” Electoral Studies 51: 93–102.
Many electoral systems constrain voters to one or two votes at election time. Reformers often see this as a failing because voters’ preferences are both broader and more varied than the number of choices allowed. New electoral systems therefore often permit more preferences to be expressed. In this paper we examine what happens when cumulative voting is introduced in two German states. Even when we allow for tactical considerations, we find that the principle of unconstrained choice is not widely embraced by voters, although in practice, too, many seem to have preferences for more than just one party. This finding has implications for arguments relating to electoral reform as well as how to conceive of party affiliations in multi-party systems.
Liam Kneafsey and Stefan Müller. 2018. “Assessing the Influence of Neutral Grounds on Match Outcomes.” International Journal of Performance Analysis in Sport 18(6): 892–905.
The home advantage in various sports has been well documented. So far, we lack knowledge whether playing in neutral venues indeed removes many, if not all, theoretically assumed advantages of playing at home. Analysing over 3,500 senior men’s Gaelic football and hurling matches – field games with the highest participation rates in Ireland – between 2009 and 2018, we test the potential moderating influence of neutral venues. In hurling and Gaelic football, a considerable share of matches is played at neutral venues. We test the influence of neutral venues based on descriptive statistics, and multilevel logistic and multinomial regressions controlling for team strength, the importance of the match, the year, and the sport. With predicted probabilities ranging between 0.8 and 0.9, the favourite team is very likely to win home matches. The predicted probability drops below 0.6 for away matches. At neutral venues, the favourite team has a predicted probability of winning of 0.7. A Coarsened Exact Matching (CEM) approach also reveals very substantive and significant effects for the “treatment” of neutral venues. Overall, neutral venues appear to be an under-utilised option for creating fairer and less predictable competition, especially in single-game knockout matches.
Media Coverage of Campaign Promises Throughout the Electoral Cycle.
A growing body of work concludes that governments fulfill a majority of their campaign pledges. However, only a minority of voters believes that politicians try to keep their promises, and many voters struggle to accurately recall the fulfillment or breaking of salient campaign pledges. I argue that this disparity between the public perception and empirical evidence is partially driven by the information voters receive in the media. I expect that only few statements on promises mention the fulfillment or breaking of a pledge, and that news outlets focus more on broken than on fulfilled promises. Additionally, I posit that partisan outlets employ more positive sentiment in statements about promises when the endorsed party holds office. Based on a new text corpus of over 480,000 sentences on political promises published in 25 newspapers during 33 electoral cycles in Australia, Canada, Ireland, and the United Kingdom, I find support for these expectations. Across the four countries, newspapers report between 1.5 and 2 times more on broken than on fulfilled promises. Tabloids, but not broadsheet newspapers, express more negative sentiment in statements on promises when the endorsed party is not in government. The results have important implications for studying negative information in mass media, election pledges, and the linkage between voters and parties.
Can voters learn meaningful information about candidates from their electoral campaigns? As with job market hiring, voters, like employers, cannot know the productivity of candidates, challengers in particular, when they elect them. The real productivity of representatives only reveals itself after the election. We explore if the information revealed during the ‘hiring process’ is a good signal of the legislative effort of elected representatives. In the incomplete information environment of election campaigns candidates should turn to credible signals to indicate their “type” to voters. Campaigns – and campaigning – are means by which candidates can, in principle, signal their motivations to voters. Is a candidate’s behaviour on the campaign trail informative about their behaviour and effort as a legislator? Does it, for example, reveal whether a candidate is more policy-seeking than office seeking? Using evidence from the European Parliament we show that campaign activity prior to the election is not related to policy-seeking behaviour in the legislature post-election. The finding also holds in two national-level settings and across a variety of measures of legislative effort. Those who campaign harder do seem more likely to win the election, but campaign effort seems to provide a poor guide to what the winner does once elected.
Retrospective and Prospective Campaign Communication. (Winner of the 2018 Manifesto Corpus Conference Best Paper Award)
Citizens attribute credit and blame to political parties and politicians when casting their ballot. However, previous comparative research has not studied how parties react to retrospective voting behavior by strategically referring to the past and present. Leveraging human coding with supervised machine learning, I uncover retrospective and prospective rhetoric in 568 party manifestos, published before 142 elections in nine countries between 1953 and 2017. Parties devote, on average, around half of their manifestos to descriptions of the past and present. Ideologically extreme parties employ substantively more retrospective rhetoric than moderate parties. Conducting separate analyses for retrospective and prospective statements, I refine two recently published studies. Incumbents and opposition parties only differ with regard to sentiment and concreteness in retrospective statements, which highlights the importance of temporal rhetoric when drawing conclusions about non-positional campaign communication. The results contribute to our understanding of representation, party competition, and responsibility attribution.
Reassessing an Established Concept Through Crowd-Sourced Text Coding.
A growing body of research analyses whether crowd workers can reproduce the ‘gold standard’ of expert-generated data. Based on the case of election pledges, I show how crowd-coding can also test for differences in perceptions of a concept between groups of experts and instructed non-experts. Comparing the most extensive reliability exercise carried out by nine pledge scholars to 3,660 codings generated by 90 crowd workers reveals considerable disagreement within and between both groups. Moreover, the carefully instructed and continuously monitored non-experts have a much broader understanding of election promises, which has important implications for analysing pledge fulfilment. The approach illustrates that crowd-coding could be used across all subfields of political science to reassess the measurement validity of a concept.
Estimating policy positions from political text is a core element in many analyses of political competition. Classical content analysis requires (costly) human experts to read and make judgements about all texts in the corpus. Benoit et al. (2016) showed that crowd workers can, much faster and more cheaply, label political texts as effectively as experts. Crowdsourced text analysis still requires judgements about every sentence in every text by multiple crowd workers, however, limiting its scalability to large text corpora. Unsupervised machine learning requires human “curation” of texts based on policy content, to allow ex-post human interpretation of results. Supervised machine learning methods leverage a relatively small training set of text labelled by humans, whether experts or crowd workers, to analyze a potentially huge volume of text out of sample, making this a much more scalable research tool. In this paper, we evaluate the effectiveness of different supervised machine learning algorithms using training sets labelled by humans, whether experts or crowd workers, to analyze both party manifestos and legislative speeches. We first replicate a very widely used left-right scale derived from classical text analysis by human experts. We then exploit the flexibility crowd sourced labels to estimate “new” policy dimensions. Our results are encouraging, suggesting that supervised machine learning based on limited training data is a viable, fast, cheap and scalable method for analyzing large political text corpora out of sample.
In coalition settings, when voters cast a ballot do they have a sense of who will form the government? The answer to this question is of relevance to questions of accountability; voters may well vote against a particular incumbent government but presumably would like some sense of what alternative will replace it. The answer is also relevant to the literature on strategic voting in coalition systems. Typically, studies of these topics rely on national level elections. But national level settings are themselves somewhat atypical electoral settings. First, these are likely to be information rich environments and provide unusually good conditions for voters to form expectations about the range of possible outcomes. In addition, studies which compare across national contexts often contain a great deal of institutional and cultural variation which muddy the focus on expectations. For these reasons, we use pre-election surveys from 19 German state elections between 2009 and 2017, in addition to three federal elections, to explore how voters form expectations in multi-party settings. We observe large variation in the ability of voters to predict actual government formation, ranging from 10 to 75 per cent. Both on the national and the subnational level we find robust evidence for ‘wishful thinking’ when predicting a government. We also find that public opinion polls heavily influence voter expectations about government formation. Our results have implications for the role of strategic voting in multiparty settings and, also, for an understanding of the ‘simple act of voting’ in complex settings.
Challenger Quality as a Moderator of the Incumbency Advantage: Evidence From a Natural Experiment in the Republic of Ireland from 1942 to 2014 (with Michael Jankowski).
We address the question of how the quality of list competitors affects the incumbency advantage. We argue that incumbents are less likely to benefit from holding office when list competitors are well-known and of high quality. To test this mechanism, we exploit the special ‘dual mandate’ characteristic of Irish local elections. Until 1999, members of the national parliament (TDs) were allowed to be represented in local and national parliaments simultaneously. By applying the regression discontinuity design to local elections between 1942 and 2014, we demonstrate that marginally elected politicians perform significantly worse when a TD competes for votes on the same list. Moreover, the incumbency advantage has become larger after the dual mandate was abolished. Our results highlight the moderating impact of high-quality competitors on the incumbency bonus.
Those Were the Days: Identifying and Explaining Nostalgia in Parties’ Campaign Communication (with Sven-Oliver Proksch).
If you would like to get access to the latest version of a paper, feel free to send me an e-mail.
I work as the Documentation Manager and Training Advisor of the Quanteda Initiative (QI), a UK non-profit organisation devoted to the promotion of open-source text analysis software, and co-author of the following R packages:
quanteda: Quantitative analysis of textual data (co-author)
readtext: Import of plain and formatted text files (co-author)
newsmap: Semi-supervised model for geographical document classification (co-author)
quanteda.dictionaries: Dictionaries for text analysis and associated functions (co-author)
quanteda.classifiers: Models for supervised text classification (co-author)
Below you can find tutorials, cheatsheets, and vignettes I have authored as a member of the Quanteda Initiative.
quanteda cheat sheet : a cheat sheet with the most important functions
Textual data visualization: plot word frequencies, wordclouds and results of text scaling models
readtext vignette: import a variety of text files into R
quanteda’s features: comparison of quanteda to alternative R and Python packages for quantitative text analysis
2019: An Introduction to Quantitative Text Analysis Using R and quanteda, University of Bergen.
2018: An Introduction to the Quanteda Package for Quantitative Text Analysis, Trinity College Dublin and University College Dublin.
2018: Introduction to Quantitative Text Analysis Using R, Methods Center of the Bremen International Graduate School of Social Sciences (BIGSSS).
2018: Introduction to Quantitative Text Analysis using Quanteda, WZB Berlin Social Science Center (with Kohei Watanabe).
2017: Data Wrangling and Visualisation Using R, Trinity Research in Social Science (TRiSS).
2018: Winner of the Dermot McAleese Award for Teaching Excellence, Trinity College Dublin.
2016: Certificate in Academic Teaching & Supporting Learning, Trinity College Dublin
University of Zurich, Department of Political Science
|09.2015–12.2018||PhD in Political Science
Trinity College Dublin, Department of Political Science
|09.2017–05.2018||Postgraduate Certificate in Statistics
Trinity College Dublin, School of Computer Science and Statistics
|09.2014–08.2015||M.Sc. in Politics and Public Policy
Trinity College Dublin, Department of Political Science
|10.2011–06.2014||B.A. in Politics and Sociology
University of Bonn, Department of Political Science and Sociology
|06.2019–07.2019||Visiting Researcher (upcoming)
Trinity College Dublin, Department of Political Science
EUROLAB at GESIS – Leibniz Institute for the Social Sciences, Cologne
|2018–||Documentation Manager and Training Advisor
Quanteda Initiative CIC
|2015–2018||Research Assistant and Teaching Assistant
Trinity College Dublin, Department of Political Science
|2014||Intern in Research Division “EU External Relations”
German Institute for International and Security Affairs (SWP)
Bonn Academy for Research and Teaching of Applied Politics
University of Bonn, Institute of Political Science and Sociology
|2016–2019||Government of Ireland Postgraduate Scholarship,
Irish Research Council
|2017||TRiSS Postgraduate Research Fellowship,
Trinity Research in Social Science (TRiSS)
|2015–2016||Postgraduate Ussher Fellowship,
Trinity College Dublin
|2011–2015||Undergraduate and Graduate Fellowship,
German Academic Scholarship Foundation
(Studienstiftung des deutschen Volkes)