Current Working Papers
- The Bundestag elections of 2009 meant a crushing defeat for the German social democrats and at the same time a low point in electoral turnout. Neither the level of turnout nor the support for the SPD has fully recovered in the last federal election of 2013. In fact, a long term comparison of changes in turnout and changes in the support for the SPD suggests that both are linked especially during the last few decades. The proposed paper tracks down this link to level of individual voting behaviour and voting intentions. Using Data from German election studies and the German General Social Survey (ALLBUS), it shows that the propensity for electoral turnout has been in decline particularly in the traditional core group of SPD voters, the members of the industrial working class. It further shows that a decline in subjective political efficacy is unlikely to be a cause of this decline, while political dissatisfaction with the personal economic situation appears to play a more important role. Finally, the paper discusses the implications of these findings for the role of social divisions in electoral behaviour.
Warum wähler ihre Stimmen zwischen Wahlen auf verschiedenen Ebenen (z.B. Bundes- und Landesebene) oder zwischen erster und zweiter Stimme in gemischten Wahlsystemen splitten, ist seit geraumer Zeit Gegenstand der wissenschaftlichen Debatte. Es wird oft angenommen, dass Stimmen-Splitting durch eine Kombination unterschiedlicher Wahlkalküle zustandekommt, z.B. der Vermeidung einer Stimmverschwendung einerseits und der Herbeiführung einer präferierten Koalition andererseits. Derlei Kalküle aufzudecken und auseinander zu halten, ist aber nicht einfach, da sie nicht direkt beobachtbar sind.
Trotz der Wichtigkeit strategischen Wählens gibt es noch keinen Konsens darüber, wie es zu messen ist. Das vorgeschlagene Papier stellt ein Finite-Mixture-Discrete-Choice-Modell für die Rekonstruktion strategischen Wählens vor, das es ermöglicht, die Kalküle unterschiedlicher Wahltypen zu spezifizieren und deren Prävalenz im Elektorat zu schätzen, sowie einzelnen Wählern Posterior-Wahrscheinlichkeiten der Verwendung der entsprechenden Kalküle zuzuordnen. Der auf diesem Modell basierende Ansatz wird dann verwendet, um Stimmen-Splitting im gemischten Whahlsystem Deutschlands zu analysieren. Basierend auf den GLES-Wahlstudien von 2009 und 2013 werden die Anteile des Stimmensplittings, die auf die verschiedenen Wahlkalküle - der Vermeidung von Stimmverschwendung und der Herbeiführung präferierter Koalitionen - sowie deren Prädiktoren herausgearbeitet.
Elff, Martin. 2015. "Challenges in Tracking Parties’ Positions in Multi-dimensional Policy Spaces by Manifesto Project data: A Dynamic Idealpoint Model and its Extensions" presented at the Manifesto Project User Conference, 4--5 June 2015, WZB Wissenschaftszentrum Berlin für Sozialforschung, Berlin
- The compilation of coded manifestos by the Manifesto Project is an invaluable resource for the reconstruction of parties political positions and their changes over time. Nevertheless, the extraction of political positions from these data is confronted by some considerable challenges. First, the relation between positions and counts or percentages -- as present in these data -- is essentially non-linear. Second, Manifesto data may reflect both political positions and the salience of policy areas. Third, the policy space wherein parties take positions is not necessarily uni-dimensional. Fourth, positions taken by parties can be expected to be autocorrelated, since parties do not invent their positions in each election ex nihilo. Fifth, it is unlikely that even the Manifesto data is free of measurement error and and that the policy topic categories used may vary in terms of their discriminatory power for policy positions. The paper points out how these problems are addressed by recently developed models of the generating process of the Manifesto data. It further proposes some extensions to these models to address the remaining challenges.
Elff, Martin and Susumu Shikano. 2014. "Bias in ML Estimation of Multilevel Models: Check the Algebra Before You Gamble in Monte Carlo" presented at the 4th Annual General Conference of the European Political Science Association, Royal Society of Edinburgh and the Royal College of Physicians of Edinburgh, Edinburgh, UK, 19-21 June 2014.
- In a recent article published in the American Journal of Political Science Stegmüller claims that ML estimation of multilevel models leads to bias and confidence interval undercoverage of coefficient estimates if the number of higher-level units is small and that this bias and undercoverage can be avoided by using methods of Bayesian inference instead. In our paper we examine the sources of bias in ML estimation of multilevel models with a small number of higher-level units. It can be shown that, for given variance components, ML estimation of multilevel model coefficients is a particular case of Generalised Least Squares, hence must be unbiased. However, ML estimation of variance parameters tends to be biased if the number of higher-level units is small. We discuss a modification of ML - restricted maximum likelihood (REML) - which reduces this bias considerably. Further, we show that the coverage error of interval estimates of coefficients found by Stegmüller can be addressed, first, by using unbiased estimates of variance parameters and, second, by using an appropriately selected t-distribution for the construction of confidence intervals. In addition to discussing the relevant literature establishing these adjustment, we replicate Stegmüller's Monte Carlo study and extend it, by taking into account REML and improved interval estimate construction. We conclude that Stegmüller claim that frequentist estimators of multilevel models are flawed is misleading and that the claim that Bayesian estimators are consequentially superior is premature.
Elff, Martin. 2014. "Separating Tactical from Sincere Voting: A Finite-Mixture Discrete-Choice Modelling Approach" presented at the Annual National Conference of the Midwest Political Association, 3-6 April 2014, Palmer House Hilton, Chicago, Illinois and presented at the EPOP (Election Public Opinion and Parties Specialist Group) Conference, 12-14 September 2014, University of Edinburgh.
: Voters' propensities to vote tactically is the main component of what Duverger calls the "psychological effect" of electoral systems. Electoral research has therefore given considerable attention to the amount of tactical voting and its consequences for electoral results. Despite the importance of the topic, research on tactical voting has not yet reached a consensus on how to measure it. The present paper proposes a finite-mixture discrete-choice model for the reconstruction of voting calculi, which makes possible to explicitly specify the calculi of different voting types and to derive individuals' posterior probabilities to have used any of these. It is further shown how maximum marginal likelihood estimates of the model parameters can be computed via an expectation-maximization (EM) algorithm. The model is applied to the case of the 2010 UK general election. It is shown that the proportion of tactical votes reconstructed from constituency data and observed choices is close the proportion of tactical votes reconstructed in terms of voters' stated motives. The paper also briefly discusses the application of the model to ticket-splitting in mixed electoral systems.
Elff, Martin, Spyros Kosmidis, and Andreas Murr. 2013. "Information Flows, Expectation Formation, and Tactical Voting" presented at the ECPR General Conference, 4-8 September 2013, Université de Bordeaux.
- Electors' propensities to vote tactically is the main component of what Duverger calls the "psychological effect" of electoral systems. Electoral research has therefore given considerable attention to the amount of tactical voting and its consequences for electoral results. In much of this literature, tactical voting has been analysed in terms of the influence of the performance of parties' candidates in prior elections, taking for granted that voters form expectations about candidates' chances accordingly. The formation of such expectations and whether they actually guide voting decisions has rarely been empirically investigated. We fill this gap by analysing the role of information flows expectation formation during the campaign for tactical voting in the United Kingdom general election of 2010. To this purpose we employ a novel approach to the analysis of tactical voting that uniquely allows to directly estimate the amount of tactical voting in a sample of survey respondents. We find that information flows indeed were relevant for the formation of expectation and for patterns of voting. Voters do use incoming information judiciously to form expectations and to vote tactically. Yet we also find that information flows lead to better estimates about the level of tactical voting than respondents' explicit statements about their expectation. A further finding is that a while a proportion of voters showed an avoidance of a hung parliament early in the campaign, that this waned after the televised leadership discussion of 15 April. Finally, we summarise the implications of from our results for information processing during an electoral campaign and for the outcome of the UK general election of 2010 in particular.
Elff, Martin. 2013. "On the Distinctiveness of Party Families" presented at the Annual National Conference of the Midwest Political Association, 11-14 April 2013, Palmer House Hilton, Chicago, Illinois.
- The concept of party families has been a useful tool for the comparison and analysis of party systems in European democracies. Central to the concept is that parties maintain an identity as member in one of the "familles spirituelles" due to their heritage from the great ideological movements of the 19th and early 20th century. It is often presumed that in a post-ideological age, party families may become obsolete. But whether this empirically is the case remains an open question. This paper examines the political distinctiveness of party families in European politics after 1945. Using a novel approach at analyzing party manifesto data it overcomes the restrictions imposed on the such analysis by the assumption of one or two ideological "super-issue" dimensions. As a consequence it is possible to establish the distinction e.g. between party families such as the christian democrats and the conservatives, which often are lumped together into a single category of "center-right" parties.
Elff, Martin and Sigrid Roßteutscher. 2014. Draft Chapter "Church Affiliation, Church Attendance and Party Support in Germany" earlier versions presented at the Third Workshop on Religion and Politics in Secularized European Democracies, 7.--8. November 2013, European University Institute, Florence, and at the First Workshop on Religion and Politics in Secularized European Democracies, 1.--3. Februar 2012, Instituto Juan March, Madrid.
- This paper tracks the development of the relation between church affiliation, church attendance and party support in Germany based on data from the German General Social Survey (ALLBUS) through the period from 1980 to 2010.
Elff, Martin. 2010. "Disenchanted Workers, Selective Abstention and the Defeat of Social Democracy in Germany", presented at the Annual Meeting of the American Political Science Association, 2-5 September 2010, Washington, DC.
- The Bundestag elections of 2009 meant a crushing defeat for the German social democrats. Never before had a party faced losses at such a high percentage in post-war Germany. The paper argues that these losses were not incurred at the margin but at the core of the social democrats' traditional constituency: Many working class citizens did not turn out to the election as a culmination of a process of growing political alienation. Data from the German Social Survey (ALLBUS) indicate that the ability to form a party preference and the readiness to participate in Bundestag elections had already been in decline in this group in the preceding years. The paper also discusses how these trends have affected the electoral outcome of 2009.
Elff, Martin. 2009. "Political Knowledge in Comparative Perspective: The Problem of Cross-National Equivalence of Measurement", presented at the Annual National Conference of the Midwest Political Science Association, 2-5 April 2009, Chicago.
- The Comparative Study of Electoral Systems (CSES) presents a unique resource for comparative research on political attitudes and behavior. From the beginning, country components of the CSES have contained each at least three items concerned with political information and knowledge. These items vary considerably across countries in terms of question format and question content. Using methods of Item Response Theory (IRT), the paper examines how these aspects impinge on the discriminance and difficulty of the items, both important aspects of their validity as indicators of political knowledge. It shows that the question content is especially important for the items difficulty: Notwithstanding the political context, items that ask for numbers (e.g. of federal states or EU member countries) are much more difficult to answer, given the level of political knowledge. Further, notwithstanding the political context, questions about foreign policy matters have a higher discriminance, that is, can better distinguish between different levels of political knowledge, than other items. The paper concludes with a discussion on how cross-national equivalence of knowledge questions can be enhanced.
Elff, Martin. 2008. "Social Divisions, Political Sophistication, and Political Equality in Comparative Perspective", presented at the Annual Meeting of the American Political Science Association, 28-31 August 2008, Boston, Massachusetts.
Although there are various conceptualizations of political sophistication and political knowledge extant in the literature, almost all of the indicators of these concepts show clear correlations with education. Further, education usually correlates with social stratification. Thus, the political presentation of the socially and/or economically disadvantaged may be hampered to the degree that political knowledge and political sophistication distorts political choices and attitudes. On the other hand, the political heuristics literature has often stressed the role of the political environment for making heuristics useful or available at all. For this reason, the same gradient in political knowledge and political sophistication in a democracy's population need not always lead to the same discrepancy in the quality of political representation of different social classes.
This paper explores the connection between social inequality, political sophistication and knowledge, and political attitudes in a comparative perspective. It those aspects of social and political systems that facilitate the use of political cues --- how party organizations, the structure of party competition, the the degree of organization of the population by trades unions, and the structure of the educational system can help alleviating the averse effects of knowledge deficiencies of certain parts of the population and thus promote political equality. The empirical analyses of the paper show that the system of government impinges on the level and the evenness of the distribution of practical aspects of political sophistication, namely of naming political issues, locating parties political positions, and knowing more or less basic political facts.