Suchergebnisse

Cover; Case Studies and Causal Inference; Contents; List of Figures ; List of Tables ; Preface ; 1 Introduction ; 2 Case, Case Study, and Causation: Core Concepts and Fundamentals ; 3 Types of Case Studies and Case Selection ; 4 Forms and Problems of Comparisons ; 5 Enhancing Causal Inference in Comparisons.

A discussion of the case study method which develops an integrative framework for causal inference in small-n research. This framework is applied to research design tasks such as case selection and process tracing. The book presents the basics, state-of-the-art and arguments for improving the case study method and empirical small-n research. Case Studies and Causal Inference delivers a self-contained and balanced discussion of the case study method on the basis of an integrative, but clear-cut framework for small-n research. It deals with case studies that develop new hypotheses for hitherto unexplained phenomena, perform classic hypothesis testing, and address puzzles in order to refine existing hypotheses. The framework further covers small-n research concerned with a range of causal effects (correlations and set relations), causal mechanisms, and the integration of both in causal explanations. By applying the framework to all steps in the research process, Rohlfing demonstrates how it can contribute to the rigor of empirical case studies and the advancement of social science theory. On the methodological side, the integrative perspective forms the basis for improving the case study method in multiple directions. Throughout the book, arguments are illustrated with numerous empirical examples from political science, public administration, and sociology.

In Qualitative Comparative Analysis (QCA), empirical researchers use the consistency value as one, if not sole, criterion to decide whether an association between a term and an outcome is consistent with a set-relational claim. Braumoeller (2015) points out that the consistency value is unsuitable for this purpose. We need to know the probability of obtaining it under the null hypothesis of no systematic relation. He introduces permutation testing for estimating the $p$ value of a consistency score as a safeguard against false positives. In this paper, I introduce permutation-based power estimation as a safeguard against false-negative conclusions. Low power might lead to the false exclusion of truth table rows from the minimization procedure and the generation and interpretation of invalid solutions. For a variety of constellations between an alternative and null hypothesis and numbers of cases, simulations demonstrate that power estimates can range from 1 to 0. Ex post power analysis for 63 truth table analyses shows that even under the most favorable constellation of parameters, about half of them can be considered low-powered. This points to the value of estimating power and calculating the required number of cases before the truth table analysis.

Accepted preprint and appendix for article published in Political Analysis as Rohlfing, Ingo (2018): Power and False Negatives in Qualitative Comparative Analysis: Foundations, Simulation and Estimation for Empirical Studies. Political Analysis 26 (1): 72-89. ; The same files plus reproduction material are available at https://osf.io/pc4df/ without restricted access, but conditional on the same requirements as listed here.

In a simulation-based analysis of Qualitative Comparative Analysis (QCA), Krogslund et al. (2015) conclude that its performance is suboptimal in several settings. I review their simulation setups and discuss three errors that were made in their analysis. First, the simulations involving inclusion thresholds are overpowered based on a misunderstanding of their role in truth table analyses. Second, the fact that a truth table analysis could exhibit model ambiguity and yield more than one model is ignored. If multiple models are derived from a truth table and they are combined into one, one overestimates the complexity of the models and underestimates their number, making it impossible to retrieve the target model of the simulation. Third, the simulations on the consequences of including irrelevant conditions intermingle sensitivity to overfitting with sensitivity to varying the inclusion thresholds. A reconsideration of KCP's simulations correcting for the errors confirms some of their findings, but also reveals that some of those errors lead to an underestimation of QCA's robustness. On a broader level, the review underscores that simulations are useful for the evaluation of QCA, but that simulation designs need to match QCA's mechanics and principles to produce valid conclusions about its performance.

The role of members of political parties is ambiguous because it entails both benefits and costs. In order to shed light on the question of whether members are an asset or a liability for parties, I examine whether parties use their ideology on a left-right dimension as a collective incentive for the appeal to actual and potential party members. A quantitative analysis of the effects of changes in membership on partisan ideological change, covering 61 parties in 11 Western democracies from the 1950s to the early 1990s, shows that there is a weak, but statistically significant, effect. An additional analysis of two mechanisms by which membership has an effect refutes the alternative explanation that positional changes of the median member account for partisan ideological change. In total, the results indicate that members are both an asset and a liability and that parties try to keep the two in balance. [Reprinted by permission of Sage Publications Ltd., copyright holder.]

The role of members of political parties is ambiguous because it entails both benefits and costs. In order to shed light on the question of whether members are an asset or a liability for parties, I examine whether parties use their ideology on a left–right dimension as a collective incentive for the appeal to actual and potential party members. A quantitative analysis of the effects of changes in membership on partisan ideological change, covering 61 parties in 11 Western democracies from the 1950s to the early 1990s, shows that there is a weak, but statistically significant, effect. An additional analysis of two mechanisms by which membership has an effect refutes the alternative explanation that positional changes of the median member account for partisan ideological change. In total, the results indicate that members are both an asset and a liability and that parties try to keep the two in balance.

The central theme of my first contribution to the symposium is the distinction between the practice and principles of social science methods (or, in the terminology of Two Cultures (chap. 1), typical practice vs. possible and best practice). The existing discussions of Two Cultures, including those in the recent symposium in Comparative Political Studies (Goertz and Mahoney 2013), emphasize the salience of this distinction for two reasons that I focus on in the following. First, I need to correct Goertz and Mahoney's (GM) potentially misleading characterization of the way in which I discuss principles of case selection in qualitative research in Case Studies and Causal Inference (CSCI). Second, in light of GM's contribution to this symposium, I should clarify and reiterate what I agree and disagree with regarding Two Cultures

In recent years, Comparative Historical Analysis (CHA) has been developed as a methodological apparatus that is distinct from quantitative research in many respects. While this is correct, it is less apparent to what extent CHA is different from and adds something to the tools and techniques known from ordinary case study research (i.e., not tied to Historical Institutionalism). As a researcher who is attached to CHA, Peter Hall is invited to elaborate on this approach. Adapted from the source document.

A fit between theory and method is essential in theory -- guided empirical research. Achieving such a fit in process tracing is less straightforward than it may seem at first glance. There are two different types of processes that one can theorise and, consequently, two varieties of process tracing. The two varieties are introduced by empirical examples and distinguished with respect to four characteristics. Failure to determine the form of process tracing at hand may lead to invalid causal inferences. Adapted from the source document.