APPLIED SOCIAL RESEARCH | METHODS SERIES | ‘Sores Esto \OWARO BIEHNAN, uy Cl, ener ery Nae = “ome. | Confi Configurational z | | «Nes tne cet oe Com j ——— - ee ethods : See aes © ROSES eer Qualitative Comparative Analysis (QCA) 0 ck ae Se mcnetnonusn and Related Techniques “Sg 2 eet enn mon etn “ esac etcane ser e aoe +s armen ORSERUATON «1, Gunurmve meseancnt O€S10 EDITED BY «Hesham = Kee BENOIT RIHOUX 0 rower coats Université catholique de Louvain Sean Suroanet Sah CHARLES C. RAGIN an Foe trod nnd etn = Beta cotter woesrons * fae - = USAC sd | © Sermsoowwin “© Soroucramere nnn | APPLIED SOCIAL RESEARCH METHODS SERIES Ps eae « acre Mente, | Volume 51 Los Angeles « London « New Doth SingaporeCopyright © 2009 by SAGE Publications, Inc. Al rights reserved, No part of this book may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without permission in weting from the publisher. For information SAGE Publications, Ine. SAGE Publications India Pvt. Lt 2455 Teller Road B 1/11 Mohan Cooperative Thousand Oaks, California 91320 Industrial Area E-mail: [email protected] Mathura Road, New Delhi 110 044 India SAGE Publications Lt, ‘SAGE Publications Asia-Pacific Pte, Lid 1 Oliver’ Yard, 33 Pekin Stret 02-01 55 City Road Far East Square London ECLY 1S Singapore 048763 United Kingdom Printed in the United States of America Library of Congress Cataloging-in-Publication Data Configurational comparative methods: Qualitative Comparative Analysis (QCA) and related techniques/[edited by) Benott Rihoux, Charles C. Ragin. cm. —(Applied social research methods series: $1) Includes bibliographical references and index. ISBN 978-1-4129-4235-5 (pbk.) 1 1. Social seiences—Comparative method. 2, Social sciences—Research— Methodology. 1. Rihoux, Benoit, 1965- I. Ragin, Charles C. H61,€59125 2009 300.72—de22 2008017858, Printed on acid free paper os 0p 0 11 1200987654321 Acquiring Edivor: ‘Vieki Knight Associate Editor: ‘Sean Connelly Edivorial Assistant: Lauren Mabib Production Editor: Sarah K, Quesenberry Copy Editor: Tony Moore Typeserer: C&M Digitals (P) Led, Proofreader: Dorothy Hoffman Indexer: Claude Rubiason Marketing Manager: Stephanie Adams Brief Contents List of Boxes, Figures, and Tables xi Acknowledgments w Introduction wit Benoit Rihoux and Charles C. Ragin 1. Qualitative Comparative Analysis (QCA) as an Approach Dirk Berg-Schlosser, Gisdle De Mew, Benoit Rihoux, and Charles C. Ragin 2. Comparative Research Design: Case and Variable Selection Dirk Berg-Schlosser and Giséle De Meur 3. Crisp-Set Qualitative Comparative Analysis (¢sQCA) Benott Rihoux and Gisele De Meur 4. Multi-Value QCA (mvQCA) Lasse Crongvist and Dirk Berg-Schlosser ‘5. Qualitative Comparative Analysis Using Fuzzy Sets (fQCA) Charles C. Ragin 6. A Commented Review of Applications Sakura Yamasaki and Benoit Rihoux 7. Addressing the Critiques of QCA Gisele De Mewr, Benott Rihoux, and Sakura Yamasaki 8, Conclusions—The Way(s) Ahead Benoit Rihoux, Charles C. Ragin, Sakura Yamasaki, and Damien Bol Appendix: Further Resources for Configurational Comparative Methods Glossary References 3 69 7 13, 7 167 19 181 185Index About the Editors About the Contributors ” Detailed Contents 207 209 List of Boxes, Figures, and Tables Acknowledgments Introduction Benoit Rihows and Charles C. Ragin Why Compare? Why Configurational Comparative Methods? Structure of This Book How to Read This Book Companion Readings 1, Qualitative Comparative Analysis (QCA) as an Approach Dirk Berg-Schlosser, Giséle De Meur, Benoit Rihoux, and Charles C. Ragin Locating QCA Epistemological Foundations ‘Small-N” Research and “Macro-Comparativ Analysis... and Beyond Key Features and Assumptions of QCA Cases and Theory Causality, Complexity, and Parsimony ‘Modest Generalization Data, Replicability, and Transparency Five Types of Uses of QCA Techniques + Notes. 2. Comparative Research Design: Case and Variable Selection Dirk Berg-Schlosser and Giséle De Meur Case Selection ‘Outcome and Area of Homogeneity ‘The “Most Similar” Versus “Most Different” Systems Designs : Further Guidance for Case Selection ‘The Selection of Conditions MSDO/MDSO: A Systematic Procedure for Matching Cases and Conditions Notes a 20 20 1 2B 25 8 23. Crisp-Set Qualitative Comparative Analysis (¢sQCA) Benoit Rihoux and Gisdle De Meur ‘The Foundation of Crisp-Se: QCA: Boolean Algebra in a Nutshell Step 1: Building a Dichotomous Data Table Step 2: Constructing a “Truth Table” Step 3: Resolving Contradictory Configurations Step 4: Boolean Minimization ‘Minimization ofthe (!] Configurations (Without Logical Remainders) Minimization of the [0] Configurations (Without Logical Remainders) Step 5: Bringing in the “Logical Remainders” Cases Why Logical Remainders Are Useful Minimization ofthe [:] Outcome (With Logical Remainders) Minimization of the [0] Outcome (With Logical Remainders) Step 6: Imexpretation Notes 4. Multi-Value QCA (myvQCA) Lasse Cronqvist and Dirk Berg-Schlosser ‘Why mvQCa? ‘Some Problems in the esQCA Example Further Limitations of Dichotomous Variables . . . and Why mvQCA Differences Between mvQCA and esQCA Notation in mvQCA Minimization in mvQCa Including Logical Remainders in the Minimization Rule Deciding Threshold Values Empirical Example: Lipset’s Theory Conclusion Notes 5. Qualitative Comparative Analysis Using Fuzzy Sets (f5QCA) Charles C. Ragin Fuzzy Sets: A Brief Presentation Fuzzy Sets Defined Operations on Fuzzy Sets 33 4 39 44 48 56 7 58 59 59 61 6s 7 co) 10 n n 4 5 76 B 83 85 87 89 89 Fuzzy Subsets Using Crisp Truth Tables to Aid Fuzzy-Set Analysis, ‘The Correspondence Between Vector Space Corners and Truth Table Rows Specifying Frequency Thresholds for Fuzzy-Set Assessments Assessing the Consistency of Fuzzy Subset Relations Constructing the Truth Table ‘The Fuzzy-Set Analysis of Necessary Conditions Application of the Procedure Analysis With SURVIVED as the Outcome Analysis With BREAKDOWN as the Outcome Conclusion Notes, 6. A Commented Review of Applications Sakura Yamasaki and Benoit Rihowx esQCa ‘Case Selection Selection of Conditions ‘Threshold Setting (Dichotomization) Dealing With Contradictory Configurations Inclusion of Logical Remainders Dealing With Contradictory Simplifying Assumptions. mvQCa Fuzzy Sets Hagan and Hansford-Bowles (2005): A Robust Application Katz, Vom Hau, and Mahoney (2008) Fuzzy Sets Versus Regression, Jackson (2005): The Importance of Case Knowledge Kogut, MacDutfie, and Ragin (2004): Informed Inclusion of “Simplifying Assumptions’ Gran (2003): Fuzzy Sets as a Typology-Building Tool Conclusion Notes 7. Addressing the Critiques of QCA Gisdle De Meur, Benoit Rihoux, and Sakura Yamasaki The Dichotomization of Data ‘The Use of Non-Observed Cases (Logical Remainders) Case Sensitivity ‘The Difficulty in Selecting Conditions 99 103 04 106 107 109 109 ut ul ns 9 120 13 123 Ra Rs 130 132 BS 136 138. 139 Bo 140 M41 12 13 44 14s “7 148 152 135 158 SS‘The Black Box Problem ‘The Temporality Problem Conelusion: There Is No “Miracle Method” Notes. 8. Conclusions-—The Way(s) Ahead Benoit Rihoux, Charles C. Ragin, Sakura Yamasaki, and Damien Bol ‘Mainstreaming "Good Practices” in Applications of QCA Connecting the Different QCA Techniques Connecting QCA Techniques and Other ‘Qualitative and Quantitative Techniques Pursuing Innovations Engaging in Collective Research Efforts ‘and Informed Methodological Debates Appendix: Further Resources for Configurationa? Comparative Methods Where to Find Further Information Glossary Note References Index About the Contributors, 159 161 64 165 167 167 169 170 m 175 179 179 181 184 188 199 209 Boxes Box 0.1 Box 0.2 Box 0.3 Box 1.1 Box 1.2 Box 13 Box 14 Box 21 Box 2.2 Box23 Box 2.4 Box 311 Box 32 Box 33 Box 34 Box 35 Box 36 Box 37 Box 4.1 Box 4.2 List of Boxes, Figures, and Tables ‘About Terminology and Labels: QCA, esQCA, mvQCA, QCA, and Software | Goals of This Textbook Pedagogical Resources in This Textbook “Multiple Conjunetural Causation” in a Nutshell Causal Relations in QCA: Assumptions ‘That Are Nor Taken Onboard Necessity and Sufficiency Back in the Picture Five Types of Uses of QCA Techniques ' [An Empirical Example Used Throughout ‘This Textbook: The “Inter-war Project” “Good Practices” (1): Case Selection in Small- and Intermediate-N Research ‘Good Practices” (2): Condition Selection in ‘Small- and Intermediate-N Research Designs “Main Steps of the MSDO/MDSO Procedure ‘Main Conventions and Operations of Boolean Algebra ‘What Is Boolean Minimization? “Good Practices” (3): How to Dichotomize ‘Conditions in a Meaningful Way Five Types of Configurations ‘Good Practices” (4): Things to Check to Assess the Quality of a Trath Table “Good Practices” (5): How to Resolve ‘Contradictory Configurations “Good Practices” (6): Four Complete Minimization Procedures to Be Run and Made Explicit myQCA Notation: Main Conventions ‘The mvQCA Minimization Rule 0 1s 2 2 4s 48 B 4‘CONFIGURATIONAL COMPARATIVE METHODS, LIST OF BOXES, FIGURES, AND TABLES si Box 43 Inclusion of Logical Remainders in mvQCA Minimization 16 ‘Tables Box 4.4 “Good Practices” (7): Threshold-Settng With mvQCA 8 “Dble3.. Raw Data Table @-Coniton Example) Br Box 45 “Good Practices" (8): Specific to mvQCA te ‘Table 3.2 Lipset’ Indicators, Raw Data (4 Conditions) 41 Box 5.1 Fuzzy-Set Membership Scores: What Are They? 90 ‘Table 3.3. Lipset’s Indicators, Dichotomized Data (4 Conditions) a Box52 “Good Practices” (9 Specific to the Calibration of Fuzzy Sets 93) ‘Table 3.4 Truth Table ofthe Boolean Configurations 45 Box 53 The Main Operations on Fuzzy Sets a ‘Table 3.5 Lipset’ Indicators, Raw Data, Plus a Fifth Condition 31 Box 54 The Criterion of Consistency in Fuzzy Sets, in a Nutshell 05) ‘Table 3.6 Lipser’s Indicators, Dichotomized Data, Plus a Fifth Condition 52 |Box 55 “Good Practices” (10 Specific BOCA ua ‘Table 3.7 Truth Table ofthe Boolean Configurations (4 +1 Conditions) 53, Box 6.1 Six Approaches to Selection of Conditions in QCA, = Table 3.8 Lipset’s Indicators, Dichotomized Data, Box 62 “Good Practices” (1!): Technical Arbitrations and Plus a Fifth Condition (and GNPCAP Recoded) sa Practical Steps Throughout the QCA Procedures es Table 3.9 Truth Table of the Boolean Configurations Box 7.1 The Importance of Case-Based Knowledge (4+ 1 Conditions, GNPCAP Recoded) ss © Open Up the “Black Box” of Causal Processes 10 ‘Table 4.1 Dichotomous Coding ofa 3-Value Traffic Light 1 Box 8:1 “Good Practices” (12): Transparency i ‘Table 4.2 Multi-Valued Scale Classifying Children’s Age ™ aoe Slee Oem = ‘Table 43 Lipset’s Indicators, Multi-Value Truth Table (4 Conditions) 80 Table 5.1 Crisp Versus Fuzzy Sets on | Figures ‘Table 2 Data Matrix Showing Original Variables and Figure 1.1 Comparative Analysis: A Typology 2 Ress Mantes Sos S Figure 2.1 Most Different and Most Similar Systems Designs is MEAS Msi of Logical AND Qntarncton) idl ‘Table 5.4 Illustration of Logical OR (Union) 98 Figure 3.1 Venn Diagram Comesponding to ‘Table 3.1 (3 Conuition Variables) 8 Table 5.5 Cross-Tabulation of Outcome Against Figure 3.2 Venn Diagram Corresponding to Table 3.4 (4 Conditions) a7 Presence/Absence of a Causal Combingtion Figwe33 Venn Diagram (5 Condlons; GNPCAP Recoded) é Table 5.6 Fuzzy-Set Membership of Cases in Causal Combinations o1 Table 5.7 Correspondence Between Truth Table Figure 3.4 Venn Diagram: Solution forthe " o 0 Rows and Vector Space Corners tos [1] Outcome (With Logical Remainders) Figure 35 Venn Diagram: Solution forthe Table 5.8 Distribution of Cases Across Causal a re estas) e Combinations and Set Taare Consistency a os) ‘of Causal Combinations as Subsets o FFigure 4.1. Data Distribution Not Allowing a Simple Dichotomization n eee " es Table 59 Distribution of Cases Across Causal Figure 4.2 Using 1Wo ‘Thresholds tor the GNFCAP Condition \ Combinations and Set-Theoretic Consistency Figure 4.3 Venn Diagrams Adapted for a Multi-Value Data Set, 82 of Causal Combinations as Subsets of BREAKDOWN 116 Figure 5.1 Plot of Degree of Membership in BREAKDOWN Against Degree of Membership in the ~D* -U*=L Combination (Comer of Vector Space) 103 nSAcknowledgments ‘This textbook isthe result ofa collective endeavor, not only because it involves seven contributors but also because it has been made possible by the support ‘of numerous colleagues, collaborators, and students around the globe. We are indebted tothe late Kriss Drass for software development. Kriss was responsible for implementation of all versions of QCA-DOS through 3.0 and for all versions of FSQCA through 0.9. Sean Davey has taken on responsibil- ity for subsequent versions of FSQCA (1.0 through 3.0), including implemen- tation of the fuzzy-set truth table procedure and the various procedures for counterfactual analysis, We thank Nancy Martin for translation work during the early stages. We also wish to mention some colleagues who have played active roles in seting up activities and further developing the COMPASS network, which has provided opportunities to develop new ideas and collaborative projects: Peter Bursens, Heike Grimm, David Levi-Faur, Axel Marx, Daishiro Nomiya, Wendy Olsen, Fritz Sager, and Geert Van Hootegem. We also thank close col- leagues and friends with whom we have debated at length over comparative methods, among them Frank Aarebrot, David Collier, André-Paul Frognier, Gary Goertz, Airo Hino, Bruce Kogut, Lars Mjoset, and Marc Swyngedouw. We also wish to thank Academia-Bruylant for its support in the publication of a first, French-language, textbook on csQCA (De Meur & Rihoux, 2002), which paved the way for this first comprehensive English-language textbook, Configurational Comparative Methods. The completion of this volume was also made possible by the support of the Fonds de la Recherche Fondamentale Collective (FRFC), through the Fonds National de la Recherche Scientifique CFNRS, Belgium), with the research grant on "Analyse de l'émergence des nouvelles institutions & parties prenantes multiples (multi-stakeholder) pour la régulation politique et sociale des conditions de travail et de la protection de Venvironnement dans des marchés globaux” (ref. 2.4.563.05.F). Work was also facilitated by a grant from the European Science Foundation (ESF), to forganize an exploratory workshop on “Innovative comparative methods for policy analysis: An interdisciplinary European endeavour for methodological advances and improved policy analysisfevaluation,” held in Erfurt from September 25-28, 2004 (ref. EW03-217). The fine working atmosphere and bibliographical resources at the Centre de Politique Comparée (Université ‘catholique de Louvain) also helped us a great dealin finalizing the manuscript. ‘We are thankful to numerous colleagues who have given us feedback in the course of the development of the FSQCA and TOSMANA software, includingwi ‘CONFIGURATIONAL COMPARATIVE METHODS Helen Giesel, Gary Goertz, Bruce Kogut, Jon Kvist, Carsten Schneider, Steve Vaisey, and Claudius Wagemarn, More specifically, we received most useful feedback on previous versions of this textbook from Simone Ledermann, Raphaela Schlicht, Carsten Schneider, Svend-Erik Skaaning, and Claudius Wagemann, as well as from {wo anonymous reviewers. We also received specific comments from Olaf Van Vliet and Maarten Vink. Finally, we owe special thanks to Svend-Erik Skaaning, who has replicated the QCA analyses ‘with some of his students at the University of Aarhus while we were finaliz~ ing this textbook—this has been most helpful in the technical consolidation of ‘Chapters 3 through 5, ‘The list of colleagues, researchers and students who have given us feedback ‘on “bits and pieces” of this textbook in conferences, seminars, and courses is naturally too long to mention, but their comments have also been valuable. In particular, we thank participants to courses we taught at the ECPR Summer ‘School in Methods and Techniques (Ljubljana), the Essex Summer Schoo! in cial Science Data Analysis and Collection (Colchester), the Oslo Summer ‘School in Comparative Social Science Studies, the Institute for Qualitative and Multi-Method Research (Phoeaix), and the Ecole d’été en méthodes quantita tives en sciences sociales (Lille). Naturally, the organizers of these venues are also to be thanked—nothing replaces such training venues when it comes to testing teaching material and new ideas. Last but not least, we thank the Sage team for their diligent support and effi- ciency, in particular Lisa Cuevas for enabling the ship to set its sails, Vicki Knight for ensuring safe arrival at the harbor after a long voyage, Sean Connelly for overall satellite guidance, Sarah Quesenberry for skilfully oper- ating the dry dock facilities, Tony Moore and Dorothy Hoffman for polishing ‘every square inch and checking every bolt from bow to stern, and finally Lauren Habib and Stephanie Adams for spreading the word around the globe, so that erowds will come visit this first vessel of its kind as it embarks on the conquest of the seven seas. On the practical side, in the team of contributors to this book, Sakura Yamasaki and Damien Bol have played a crucial role in the maintenance of the COMPASSS bibliographical database out of which the list of references has been extracted, All this being said, we fiercely claim intellectual property on any remaining error, approximation, or omission. Introduction Benoft Rihoux Charles C. Ragin WHY COMPARE? WHY CONFIGURATIONAL, COMPARATIVE METHODS? Comparison lies tthe Rear of human reasoning and is lays there in he bservation of the worl—"hinking without eomparion Is unhinkale™ (Swanson, 1971, p48). Inded, even he shuervtion of singular phenomena is empty if we donot engage in comparison: A phenomenon o object ar Be identified ay such only it resognized as diferem fom eter phenomenor objets (Arebot& Baa, 2003). For instance, we know that apples a ot pears because we have compared the wo. More specifically comparion i key operation in any empiilstenfc efor Theres longline of soars who bate efetedupon isan ped thsempicaly—all the way fom Arist probably the eunde of gts comparative approach) to de Toequville, Weber and Durkin are contemporary works by Sartor 1970, 1991, Liphat ep. 1971) Marat (1988). For one hing, any deserve efor ay plogy orcas cation involves compariaon (Bailey, 199) To consider both ape sl pes belonging othe etegey of “rity” we must compare“ and sno fur” inthe broader category of "plan" and oon. One we have defined thee ory of fruts" we can come tote consi hat an orange is aoa by Comparing some ey popes ofan orange with hse ofan pple an of apse While also ting fut an crane shares sone specific characteristics with lemons and grapetuis. Thus oranges and lemons, onthe oe hind nd pes and pear o the ther hand, belong to vo diferent byes of fats, Ofcourse, sich everyday comparisons may seem iv and indeed many of those menial operations remain impli in our reasoning The purpoe of this textbook ito demons hat comparisons abst and poweral mone tl operation, canbe taslted int ast of systematic comparative metals and eenique. Although ths volume isnot abou fas, we sal Jemonstte that such methods and ehiigues va indeed be Mul applied ay Aisciptnes, inthe soil seienes broadly defined and aso beyond Systematic comparison ist ey operation inal experimestal and natal scenes For instance, we know ht waters oling when is eta 0100 because we have compared the ite of water below 100°C, a 100°Cs adwa CONFIGURATIONAL. COMPARATIVE METHODS above 100°C, while controlling for contextual parameters such as atmospheric pressure and altitude above sea level. Incidentally, because we are able to Control all these contextual parameters, and because we are able to manipulate fone specific condition—temperature—we are able to demonstrate that @ change in temperature actually causes the water to boil. This is why the expet- imental sciences are able to make such strong and simple causal statements. Yet, in most social and behavioral sciences, perhaps with the exception of some branches of psychology, real-life laboratory-like experimentation is neither empirically possible nor ethically desirable. To pursue our example further: In antiquity, slaves were used to scientifically demonstrate that water that is brought to a boiling temperature also happens to bum human skin, by ‘comparing the effect of plunging the hand of a slave into lukewarm, as com- pared to boiling, water. And Cleopatra used slaves to examine the more or less lethal effects of various poisons in her apples and oranges. Obviously, con- temporary social scientists cannot do this, nor would they want to. This is where the comparative method comes into play. Itcan be considered a crude substitute for experimentation (Lijphart, 1971): We observe empirical phenomena—analytical units, “cases” (Ragin & Becker, 1992}—while also ‘controlling for contextual conditions (see Chapter 1). In social science, those ceases, as we know, are intrinsically complex, multifaceted, often with blurred boundaries. This is why “thick,” single-case studies have always played an important role in many disciplines: They allow a deep understanding of that single case, The main limitation of single-case studies, however, is that it is very difficult to engage in any form of generalization, as the key findings and conclusions are mostly limited to that single case. How can one compare these complex cases? During recent decades, an increasing number of social scientists have been opting for multiple case stud- ies as a research strategy. This strategy aims at meeting the need to gather in depth insight into different cases and to capture their complexity, while stil attempting to produce some form of generalization (Ragin, 1987). It also coin~ cides, during the last few years, with renewed interest in case-oriented research (eg, George & Bennett, 2005; Gerring, 2006; Mahoney & Rueschemeyer, 2003), Such @ strategy is also adopted because many relevant and interesting objects are “naturally” limited in number: nation states or regions, political crises, wars, firms of a certain type, and so on. These are naturally limited, or “cmall-N" (or “intermediate-N"—see Chapter 2), populations of cases. In many instances the (ex post) comparison of case study material is rather 1: not formalized. The major ambition of the methods and techniques presented in this textbook is to allow systematic cross-case comparisons, while at the same time giving justice to within-case complexity, particularly in sinalt- and intermediate-N research designs “Too: INTRODUCTION ix ‘The cover heading for all these methods and techniques is Configurational (Comparative Methods (CCM). In a nutshell, this heading indicates that in order to enable the systematic comparative analysis of complex cases, those cases ‘must be transformed into configurations. Simply said, a configuration is a spe- cific combination of factors (or stimuli, causal variables, ingredients, determi- nants, etc—we call these conditions in CCM terminology) that produces a given outcome of interest. As shall be explained at length in the next chapters, the conditions will be envisaged in a combinatorial way—hence enabling one to model quite a high level of complexity even with only a few conditions (One key question we shall address is the following: Which conditions (or combinations thereof) are “necessary” or “sufficient” (or possibly both neces- sary and sufficient) to produce the outcome? Ina non-formal way (for more on this, see p. 10, Box 1.3; see also Caramani, 2008), let us say at this stage that: ‘© A condition is necessary for an outcome if i is always present when the outcome ‘occurs. In other words, the outcome cannat occur in the absence ofthe condition. ‘© A condition is sufcien for an outcome ifthe outcome always occurs when the condition is present. However, the outcome could also result fom other conditions. For instance, holding competitive elections is a necessary condition for a state to be considered democratic. However, itis not a sufficient condition because comprehensive civil liberties must also be present for a state to be considered democratic. Nonetheless, the absence of competitive elections is a sufficient condition to qualify a state as non-democratc, as a democracy can- not exist without competitive elections." Under the heading of CCM, we place four specific techniques: Qualitative ‘Comparative Analysis using conventional, crisp sets (¢sQCA, often simply labeled QCA in the literature), multi-value QCA (mvQCA),fuzzy-set QCA (f8QCA), and MSDOMDSO (most similar, different outcome/most diferent, same outcome). cere Pacem e “We use the label QCA as an umbrella term that captures the three main “gps (Boolean lave and hazy set) aa group Afr ll hey share ‘many commonalities (see Chapter p. 1). (Continued)~ CONFIGURATIONAL COMPARATIVE METHODS (Continued) "+ When referring explictly to the original Boolean version of QCA, we use Y). ‘Several diferent combinations of conditions may produce the same outcome (AB + CD 3+ indicates a Boolean or). Depending on the context, 2 given outcome may result from a condition whan itis present and also when itis absent (AB > Y but also aC >). In this example, [A] combined with [B] produces the occurrence’ of the out- come, but its absence {3] combined with (C] also produces the outcome. In other words: Different causal “paths"—each path being relevant, in a /—may lead (0 the same outcome (De Meur & Rihoux, 2002). The term “multiple” refers to the number of paths, while the term “conjunctural” conveys the notion that each path consists of a combination of conditions. Thus multiple conjunctural causation contains the notion of equifinality, which simply means that different paths can lead to the same outcome. It should be noted that this runs completely against key assumptions on which ‘mainstream statistical techniques rest—for example, additivity, meaning that agi across all cases, regardless of the values of other causally relevant conditions, (Schneider & Wagemann, 2007, forthcoming). ‘Since it views causality as context and conjuncture specific, QCA rejects any form of permanent causality (Ragin, 1987). This isin ine with the earlier works of IS. Mill (see p. 2), Bottom line: By using QCA, the researcher is urged not to specify a single causal model that best fits the data, as one usually does with statistical techniques, but instead to determine the number and character of the different causal models that exist among comparable cases (Ragin, 1987). distinct wa en factor is assumed to have the same incremental effect on the outcome (QUALITATIVE COMPARATIVE ANALYSIS (QCA) AS AN APPROACH, ° It thus goes beyond the (often superficial or misleading) means, correlations, and regressions—computed across all cases at the same time—which average cou the respective constellations and ignore specifi, distinct patterns and “out= ” (see, e:g., also Berg-Schlosser & Cronqvist, 2005; Berg-Schlosser & Quenter, 1996; Ragin, 2006a).* Actually, with QCA, if a given combination of conditions “explains” only one single case, itis not a priori considered as less relevant or less important than another combination of conditions that would account for, say, 10 or 15 eases—because each case matters in most applications of QCA (see pp. 23, 155). In this sense, QCA moves away, quite radically, from simplistic, probabilistic causal reasoning (De Meur & Rihoux, 2002); in its case- orientedness, it is more geared toward diversity (Ragin, 2006a). ‘Thus QCA broadens the usual frame in the analysis of causality, by relaxing several common assumptions, First, “additivity” is no longer assumed: This means thatthe idea that each single cause has its own separate, independent impact on the ‘outcome is abandoned and replaced by the assumption that “conjunctural causa- tion’ is at work, meaning that several causes can be simultaneously present (or be combined, somehow), constituting a “causal combination,” for the outcome {0 ‘occur, Second, a given causal combination may not be the only route toa specific result; other combinations also may be able to produce it. Third, the uniformity of ‘causal effects is not assumed; on the contrary, a given condition may, combined ‘with different others, sometimes act in favor of the outcome, and sometimes, differ ently combined, act agains it. Fourth, causality is not assumed to be symmetrical — rather, causal asymmetry is assumed, meaning that the presence and the absence of the outcome, respectively, may require different explanations. Cnr) Causal Relations in QCA: ‘Assumptions That Are Not Tal it is eruclal to bear in mind that QCA does not take onboard some basic assumptions that lie at the heart of the mainstream statistical approach (and , ‘thus underlie most statistieal techniques).In QCA: + Permanent causally s ot assumed. 1 Uniformity of causal effets isnot assumed. Unie homogeneity is not assumed. Adktivy isnot assumed. ‘© Causal symmetry isnot assumed. Note that other core mainstream statistical assumptions, such as linearity and 40 0n, are not taken onboard eitherw CONFIGURATIONAL COMPARATIVE METHODS Of course, QCA techniques do not guarantee the final grasp of the “true” causal grounds of a given phenomenon because the issue of causality is a ‘much more complex matter (see, e.g., Abell, 2004; Gerring, 2005; Mahoney, 2004). Besides, the conclusions of any empirical analysis (QCA or any other) are totally dependent on the choice of “ingredients” put under the microscope, including the condition variables as they have been operationalized as well as the selection of cases. Yet, if several competing theories try to explain the same result, QCA techniques will quickly disqualify the theories that are unable to disctiminate correctly between cases with and without the outcome under study, This will be indicated by the presence of so-called contradictory con- figurations (see p. 44) Among the remaining theories, those that best satisfy the “parsimony prin- ciple” (Occam’s “razor") will emerge. The parsimony principle, successively reinvented and reinforced through the centuries, can be translated into the com- ‘monsense adage: “Why make ccmplicated when one can make simple?" Or, as Einstein put it in his famous dictum: One should express things “as simply as Possible, but no simpler” To sum up: QCA techniques strive to achieve some form of “short” (parsimonious) explanation of a certain phenomenon of inter- cst, while still providing appropriate allowance for causal complexity. ‘Naturally, the search for causal regularities implies the acceptance of the postulate that there are indeed underlying causal regularities in human and social phenomena, even if sketchy (Ragin, 1987, p. 19; Skocpol, 1984, pp. 374-375: Zelditch, 1971). In QCA, as discussed in Box 1.3, two key regularities are framed in terms of necessity (“necessary [combinations of] conditions") and suffi- ciency (“sufficient (combinations of] conditions”). In fact, any empirical scien- tific process—even those within the “hard” seiences—is based on this postulate. The opposing postulate, that of an “unstructured chaos” of phenom- cena, would preclude the search for explanations as well as for meaning. Box 1.3 sity and Sufficiency Back in the Picture” [Note that the key concepts of necessity and suficiency (as defined on p. xx) are very much Inline with the mutiple conjunctural view on causation, Indeed, a given path toward an outcome usually consists in a combination of conitions ‘that is sufficient (a sufficient combinetion or “intersection” of conditions) to pro= luce that outcome. However, this path is not always necestary, a8 some other alternative paths (with diferent conditions, a least party) could very well pro duce the same outcome. Let us pursue the example used on p.xix and consider ‘three possible conditions leading to the outcome building a democratic state": (QUALITATIVE COMPARATIVE ANALYSIS (QCA) AS AN APPROACH, un * Condition A: Hold reguar competiive elections. ‘+ Condition B: Ensure comprehensive civ liberties. ‘+ Condition C: Ensure independence of poltical decision-makers vis-ivis the miltary leadership. In this example, chere could be two paths leading to the outcome of interest: ‘© Path 1: the combination of A and B ‘= Path 2:the combination of A and C ‘This can be translated as follows: ‘Path 1 isthe firs suicent combination of cndiions leading to the outcome. ‘Path 2s the second sufcien combination of condos leading tothe outcome. [Neither ofthese ewo paths, considered separately is both sufficient and nec- ‘essary (as there is alvays an alternate path leading to the outcome). Noce, finally that one condition (A: Hold regular competitive elections) is present in both paths, Hence, we can say that: ‘+ Ailsa necessary condivun for the outcome to occur (because It is always present when the outcome occurs). ‘+ However i isnot a suficient condition, because condition A alone does not Produce the outcome—ic needs to be combined with ether condition B or C. Modest Generalization Generalization is an important part of any empirical scientific endeavor, ‘The goal of research is not limited to description, as exhaustive as possible, of some corpus of observations, The search for “explicit connections” (Ragin & Rihoux, 2004a) or “specific connections” (see Rihoux, 2008b) takes an impor- tant part in the process of understanding. Explicit connections give a formal shape to observed regularities that occur in the data set, and this allows for further investigations, as they are dissected to elaborate an “explanation” — an attempt to describe the mechan 4m at work. They also give a predictive tol, providing assertions on the behavior of new, not yet observed cases and there- fore offering an opportunity to test the model and go a step further, ‘Without the ambition to generalize, in the search for explanations, research ‘would produce only tautologies and descriptions, This is not to say that more2 CONFIGURATIONAL. COMPARATIVE METHODS. interpretive or “thick” descriptive work is devoid of value—indeed such work can yield very useful insights to grasp phenomena, to understand their deeper ‘mechanisms, to gain an understanding of complex cases (Gerring, 2006; Ragin & Becker, 1992), But it is crucial to recognize the importance of producing new conjectures and to take the tisk of confronting them with new data, The degree of maturity and robustness of a generalization will strongly ‘depend on the quality of the empirical data set constructed by the researcher, ‘and it will generally be a long and hard job to produce it, with many trials and errors, new questionings, and assessments. Contrary to popular myth, those readjustments should not be considered opportunistic manipulations of data; they are necessary steps in their elaboration as researchers increase their sub- stantive and theoretical knowledge—this is why QCA techniques, as shall be ‘demonstrated in further chapters, are iterative by nature. ‘A good index of the quality of research results could be precisely their ability to withstand refutation when confronted with new cases. In this respect, wwe should remember that a theory maximizes its robustness when it avoids individualizing explanations—that is, when it avoids providing a specific “explanation” for each specific case (it is then only an accumulation of “descriptions,” and not an “explanation”). Only generalization makes it possi- ble to achieve more succinct explanations—such as in the example in the pre- vious section, where condition A is identified as a key regularity (a necessary condition). This again speaks strongly in favor of parsimony. Yet, the efforts toward generalization that have a reasonable chance of success must stay inside the initial “homogeneity space,” within which the empirical data set is contained (see p. 20.). Nothing supports the idea that con- ditions not included in the analysis would not affect the results ofthe analysis. Hence a well-executed QCA should go beyond plain description and consider “modest generalizations": QCA results may be used in support of “limited his- torical generalization” (Ragin, 1987, p. 31). More specifically, from a system- atic comparison of comparable cases, itis possible to formulate propositions that we can then apply, with appropriate caution, to other similar cases—that is, ‘cases that share a reasonable number of characteristics with those that were the subject of the QCA. Note that this view on generalization is much more modest than statistical inference, which allows very broad generalizations (from a sample of, say, 1,000 respondents to a population of millions of individuals) Data, Replicability, and Transparency ‘As mentioned above, QCA techniques require that each case be broken down into a series of features: a certain number of condition variables and an outcome variable, For instance, if we consider athletes as cases, if the outcome is the t i [QUALITATIVE COMPARATIVE ANALYSIS (QCA) AS AN APPROACH 8 ability to throw a discus beyond 60 meters, then some conditions could be being. tall (versus not tll), being fast (versus slow), being muscular (versus thin), and. so on, Then we could measure these attributes for each “case” (athlete): Case 1 ‘could be tll fast, and muscular; Case 2 not tall, fast, and thin; and so on. ‘This means that, as with statistical analyses, QCA techniques allow one to develop an analytical strategy. However, this segmentation into variables does not affect the perception of each case as a whole. The aim here is to allow for ‘major concerns of both quantitative (defining variables) and qualitative (keep- ing in touch with the holistic perspective) approaches. Having done so, one will be able to compare cases as “whole units,” each one of these being defined as a combination of features (i.e., as a configuration, as defined p. xix) In this analytical process, QCA techniques enable us to take into account both “qualitative” and “quantitative” phenomena. When the first QCA tech- nique (csQCA) was developed, Ragin, and several other scholars, used the “qualitative” label to refer to phenomena that vary by kind rather than by degree and also to stress the importance of considering cases as specific and complex configurations, or combinations, of features (Ragin, Berg-Schlosser, & De Meur, 1996, p. 749). Incidentally, we should mention here that itis per- feetly possible to work with “subjective” or “qualitative” data using QCA. The only practical requirement is to be able to transform these data into categories fr numbers, For instance, if our cases are political parties after a given elec~ tion and a condition is defined as “perception of electoral defeat by the party activists,” the nature of the data could be very subjective (ie, based on discussions with party activist, assessment of the “tone” in the party press, a feeling about the atmosphere in the post-election party congress, using partic~ ipant observation methodology, etc.). For any given party, one could still allo- cate a numerical score of “I” (“yes, or mostly yes”) or “0” (“no, or mostly no") ‘on this condition. However, csQCA, as well as the other QCA tethniques, is also able to consider phenomena that are “more” or “less” similar—that is, to express a degree of (dis)similarity (the differences therefore being of a quantitative nature)—as well as phenomena that differ by their nature (the differences being in this case qualitative), For instance, a “wealth versus poverty” variable could bbe based:on some fine-grained data on yearly income of individuals—this ‘would then be atypically “quantitative” measurement of wealth versus poverty. ‘Alternatively, one could consider—for instance, following some official UN ot OECD criteria—that the crucial analytical difference is to consider whether a given individual is “poor” or “not poor” Usually, this is done by specifying an income threshold (say, below 1,000 EUR/month, in a country such as Germany) below which one is considered “poor” This is exactly what will be done in the basic, dichotomous esQCA: switching, in the course of the analysis, from1“ CONFIGURATIONAL COMPARATIVE METHODS quantitative (numerical) to dichotomous (still numerical, but tapping a more ualitstive distinction) variables, using substantive knowledge to guide the dichotomization. By dichotomizing the originally fine-grained (quantitative) data, cur aim is to identify what can be considered a more fundamental, more basic distinction, of a more qualitative nature. In this process, similarly to the quantitative (mainly statistical) approach, QCA offers tools that are both formalized and replicable. QCA techniques are formalized in the sense that they are based on a particular language (Boolean algebra and set theory) whose rules and well-defined solutions for- ‘malize and translate the rules of logic (formal operations and algorithms; see Chapters 3 to 5). Because these formal rules are fixed and stable, they allow replicability. Simply stated, this means that another researcher using the same data set and selecting the same options will obtain the same results (King et al., 1994, p. 26). This is a major asset of QCA techniques compared. to many ad hoe of less formalized qualitative techniques. From a certain perspective, one could say that replicability provides the “scientific” charac- ler of the approach, in the sense that it eliminates vagueness and interpreta- tion in the application of techniques (mathematics, for example, is universal as far as technique is concerned). Another advantage of QCA techniques is their transparency, They demand that the researcher, at several points in the analysis, acts with transparency in his or her choices—selecting variables, processing them, choosing tools for the analysis, intervening during the analysis, and so on. During this process, the researcher regularly refers back to the cases with all their richness and speci- ficity. This back-and-forth “dislogue with the cases,” combined with the trans- parency of choices, is unquestionably a virtue of QCA techniques. What also makes this transparency possiole for QCA techniques is that the formal lan- ‘Buage used by the software takes its inspiration from principles used in every- day life and, for this reason, cen be mote easily understood by non-specialists With most statistical tools, the researcher enters the data and the software finds the “solution.” In contrest, QCA opens the “black box” of formalized analysis, by demanding from researchers not only that they make choices but, also that they account for them. Using QCA, researchers must be engaged in the analytic process, for it is not mechanical or “push-button.” For sure, such, requirements should also apply to statistical work. The difference, with QCA, is that the user is more active, gets a better grip on the “mechanics” of the formal operations,* makes more decisions in the course of the analysis, and follows an iterative logic, with frequent “returns to the cases.” Researchers may feel uncomfortable with this, but this lack of comfort is beneficial, because it compels them to use critical thought during the (QUALITATIVE COMPARATIVE ANALYSIS (QCA) AS AN APPROACH Is analysis and opens up the researcit to others for confirmation or falsification (Popper, 1963). FIVE TYPES OF USES OF QCA TECHNIQU QCA techniques may be exploited in at least five different ways. According to their specific needs, researchers can use different features af QCA. Here, we consider only the three versions of QCA (esQCA, mvQCA. fsQCA). MSDO/MDSO is a technique geared toward one specific type of usage. Coaey ed kee Summarizing data ‘Checking coherence of data Checking hypotheses or existing theori ‘Quick rast of conjectures Developing new theoretical arguments First, QCA techniques may be used in a straightforward manner, simply to summarize data, to display them in a more compact way, and to deseribe more synthetically the relevant empirical universe. This is thus a purely descriptive use of QCA. More specifically, this is done by means of using the software to generate a synthetic table that shows, in a straightforward way, how some ccases cluster together—the so-called truth table (sce Chapters 3 to 5). In this \way, the researcher will be able to bring to light similarities between cases that may, a first sight, seem quite different. QCA is therefore an excellent tool for data exploration, 1 Second, the researcher may take advantage of QCA to check the coherence of the data. During the analysis, one often detects contradictory configura- tions—that is, cases that are identical with respects to causal conditions, but different in outcome (see p. 44). Contradictions are plainly displayed in the truth table produced by the software. Detecting e: rot necessarily mean that researchers have failed. On the contrary, contradic- tions wil tell them something about the cases they are studying. By seeking a solution to these contradictions, the researcher will both get a more thorough radictions, however, does6 (CONFIGURATIONAL COMPARATIVE METHODS knowledge of relevant cases (through his or her “dialogue withthe cases”) and develop @ more coherent body of evidence (see p. 48). Third, QCA may be used to fest hypotheses or existing theories. More pre= cisely, it enables us to corroborate or falsify these hypotheses or theories. When using QCA in this way, the researcher aims at operationalizing some theory or hypothesis, as explicitly as possible, by defining a series of condi- tions that should yield a particular outcome. QCA is @ powerful tool for this kind of application because it allows theory testing or hypothesis testing that is both systematic and empirical. When the researcher discovers, through QCA, a large number of contradictory configurations, it may enable him or her to falsify the theory or hypothesis (a very important achievement from a Popperian perspective; see above and p. 50). Furthermore, QCA allows us to refine the hypothesis testing process, by taking into account the actual number of cases related to falsifistion or corroboration AA fourth use, close to the former one, is the quick test of any conjecture formulated by the researcher him- or herself—that is, without testing a pre- existing theory or model as a whole. This is another way of using QCA for data exploration. The researcher specifies an expression (a formula) reflecting 1 specific conjecture, for example, to test an ad hoe theory or part of a theory. This yields a truth table, which allows the researcher to check whether his or her conjecture was aecurate—whether it is confirmed or falsified by the set of ceases under analysis, Last but not least, QCA also may be used in the process of developing new theoretical arguments in the form of hypotheses. By obtaining a truth table free of contradictions and then conducting QCA, the researcher obtains reduced expression (called a “minimal formula”). This may then be interpreted through a “dialogue with the cases" to yield new theoretical arguments. In this way, QCA can be used in a more grounded manner. (One specific technique is especially relevant for this fifth use of QCA. It consists of revising by hand the reduced expression (results of truth table analysis) generated by the software. More specifically, the researcher treats these results as a conventional algebraic expression (Boolean sums of products) and factors it to highlight shared conditions, or to rearrange it in other algebraically acceptable ways so that it speaks as directly as pos- sible to theoretical and substantive interests (see, for example, p. 58). As such, however, QCA does not yield new theories. What it may do, once performed, is to help the researcher generate some new insights, which may then be taken as a basis for a further theoretical development or for reexamination of existing theories. Only by returning to empirical cases will it be possible to evaluate whether it makes sense to highlight a partic ular condition, (QUALITATIVE COMPARATIVE ANALYSIS (QCA) AS AN APPROACH, 0 ei Ses GT EERE GCA Ri ki Le ey SYA BS clio ah Tes eee Ste aa rr es coe roe ey Ea Gnawa mines Se Aa ca aby cami on areca oe ee een oe erro (eure Soi sere oats See ee ee aes ree cate Cail prathapml ‘of conditions, and diversity. pease es ee neces con noes goes soy retin eae een Se ene ee on cy ns encase awe Fase Si eblai Woden bar Key Complementary Readings ‘Goerez (2006), (1843) Popper (1963),Ragn(1987,2000,2003, 2006). NOTES 1. Fora complementary view, see Caramani (2008). 2. And also because ofthe “limited diversity” problem (see p. 27) 3. A note of nuance: One may attempt to empirically and systematically test some fragments of such grand theories—see for instance how Andersen (2005) succeeds in testing Lahmann’s systems theory, Part of Bourdie's own work has also consisted in deriving concrete and testable empirical propositions from his grand theoretical framework—but alas only very few of his contemporary “believers” have followed that path. 44. See p. 34 and Box 3,1: Main Conventions and Operations of Boolean Algebra 5. Ibid. 6. tis far to say that the broad majority of mainstream statistical work does litle to deal with causal complexity. However, some suggestions have been made in the6 CONFIGURATIONAL COMPARATIVE METHODS Statistical literature as to how t0 deal with issues such as asymmetrical causation, necessity, sulficieney, and so on. See Schneider and Wagemann, 2007, forthcoming. 7, Example suggested by Lasse Crongvis, further elaborated by Benoit Rihoux. extensive discussions, see Goertz, 2006b, Schneider and Wagemann, 2007, wgemann & Schneider, 2007 8, These logical operations. 3s well as all their underlying assumptions, are much simpler than statistical operations. Very few users are really able to understand the deep, ‘mechanisms and assumptions st work behind the statistical operation. Comparative Research Design Case and Variable Selection Dirk Berg-Schlosser Gisdle De Meur atone ‘After reading this chapter, you should be able to: © ery earth re f breton win which you sl your ‘Sses wih ae precsipaton inthe cnon ofthe aun of eres + Choose the type of research design that most sued to your needs: “Sos sa” yrs dein "most lren” systems Geen, oF 8 combination of both + Make wel-iformes choices onthe exact numberof cates tht wil be ined nthe ani. on methodlogal and prac grounds + Procaad to model buldng—namely deny poten! condons to be Inchded in the. model), reying on case-baced knowiedge and theoretal considerations . + Gasp the ey gas and supe of SDOMOSO.a ford procedure Shar among oer things makes ie posable to decrease the number of Centtione when rand” condtons are foo numerous to proceed wth QCA Each empirical field of study can be described by the cases (“units”) analyzed, the characteristics of cases (“variables”) being considered, and the number of times each case is observed (“observations”) (King, Keohane, & ‘Verba, 1994, pp. 51 ff.). We focus especially on the first two aspects, cases and variables. For many analyses it is indispensable to examine variation over20 CONFIGURATIONAL COMPARATIVE METHODS. time, and such longitudinal study can be conducted using both quantitative (eg. Petersen, 1993) and qualitative (¢.., Griffin, 1993; Heise, 1989) methods. However, studies of this type pose their own special problems (sce p. 161), which are not dealt with at this point and are not the key focus of the research designs and techniques discussed in this chapter. In case-oriented, small- and intermediate-N situations, which is the domain ‘we are mostly concerned with here (see pp. 3-5), both case selection and vari- able selection are essential for the further steps of inquiry. Both should be guided by explicit theoretical concerns, but, atleast initially, they also may be exploratory in nature, starting at a relatively low level of theory building (“ana- Iytie induction,” Blalock, 1984), Only in later stages can the range of explana tions be determined and systematically tested. In any event, it is of crucial importance before engaging in QCA proper (through esQCA, mvQCA, fsQCA) to develop a specific comparative research design. Naturally, developing earch design covers many other aspects than simply case and variable selection—but we choose to concentrate on these two operations because they are particularly crucial in comparative research designs. CASE SELECTION Outcome and Area of Homogeneity At the outset of any investigation an area of homogeneity, a “domain of investigation” must be defined that establishes boundaries within which cases are selected, Cases must parallel each other sufficiently and be comparable along certain specified dimensions. This is the meaning of the common adage that “apples and oranges” should not be compared. In this regard, the subject ‘matter and the problem we are interested in (the outcome, in QCA terminol- ‘ogy) must first be specified to make any sense. Thus, these fruits may well be compared concerning theit sugar or water contents, their nutritional value, and so on, provided that some other dimensions on which they share some ‘common properties (¢.g., they both grow on trees, they are both suitable for human consumption) have been made explicit. The specification of relevant ceases at the start of an investigation thus amounts to an explicit or implicit hypothesis that the cases selected are in fact alike enough to permit compar- isons, In other words, the cases must share enough background characteristics, which in turn can be considered as “constants” in the analysis. Thus, the primary consideration in delimiting cases for a small- or intermediate-N com- parative study is the outcome. (COMPARATIVE RESEARCH DESIGN 2 cere Rae ase ec) Re aoe ae ecca a From here onwards, and all the way through this chapter and the three fol- lowing (Chapters 3 to 5), we shall use, as 2 “realife” example, a large-scale ineernational research project on the survival or breakdown of democracies In Europe in the interwar period (the “incer-war project” Berg-Schlosser & ‘Mitchell, 2000, 2003). In the inter-war project, the breakdown or survival of democratic regimes in inter-war Europe—the outcome the research team was striving to explain— prestupposes the prior existence of some form of democracy in the selected ceases. In addition, some limitations in time and space can also enhance the homogeneity and thus the comparability of the cases examined. For example, certain kinds of colonial or other forms of external domination or religious: cultural influences may be useful criteria for selecting a specific group of ‘cases, A key point to remember is that a clear definition of the outcome of interest must be explicated ata very early stage of the QCA, because itis indis- pensable for the selection of cases, ‘A second consideration concerns the extent of diversity within the selected universe. In this regard, a maximum of heterogeneity over a minimum number of cases should be achieved. In the inter-war project, for instance, both \l breakdowns of democracy should be considered, and among the latter perhaps some more specific variants such as fascist versus more gener- ally authoritarian outcomes. Generally, itis advantageous to include cases with 1 “positive” and cases with a “negative” outcome.” ‘The “Most Similar” Versus “Most Different” Systems Designs Once the universe of investigation and the outcome of interest have been clearly identified, two opposite strategies now become possible. One is the ‘most similar, the other the most different systems design. ‘These have been explicitly formulated and discussed by Preeworski and ‘Teune (1970)—these authors use the term system to designate a complex case. ‘The most similar systems design is “based on a belief that a number of theo reti¢ally significant differences will be Found among similar systems and that these differences can be used in explanation” (p. 39). By matching these SeRpaDnnETE RIE e Nn vernrtret Tw)2 CONFIGURATIONAL COMPARATIVE METHODS similar cases as much as possible, most of the variables can be “controlled.” Mill's (1967 [1843}) “indirect method of difference” (see p. 2), where differ- ent outcomes may be attributed to the remaining factors that differentiate these ceases, now becomes applicable. Even though only rarely just a single factor will remain to which the effect can be attributed, at least many others can be excluded, and the remaining ones can be examined more closely in a theore- tically guided qualitative manner. The “internal validity” of the observed rela- tionships thus can be greatly enhanced (see also Cook & Campbell, 1979) ‘The opposite strategy, the “most different” systems design, “seeks maximal heterogeneity in the sample of systems, {and] is based on a belief that in spite of intersystemic differentiation, the populations will differ with regard to only a limited number of variables or relationships” (Przeworski & Teune, p. 39). ‘This “contrasting” of cases thus eliminates all factors across the observed range that are not linked to an identical outcome. In this way, more “univer- sal” explanations are sought as far as the selected area of homogeneity is con- cerned, To some extent, thus, the “external validity” of some hypothesized ‘causal relationship (Which have to be identified at the disaggregated level, not, at the system level) can be extended and the range of its applicability includ- {ng certain limitations in time and space can be established. ‘These designs can be visualized in a simple manner showing the respective intersections for a systematic matching and contrasting of cases. Consider Figure 2.1, which visually represents the contrasting emphases of MSDO (most similar [systems with a, different outcome) and MDSO (most afferent {systems ‘with al, similar outcome). In this figure, each circle represents a case, and the intersections represent their commonalities. In an example of three cases that are, in one instance, “most similar” but have a different outcome (MSDO) the commonalities of the cases are indicated by the white area, whereas the shaded ‘areas indicate their remaining idiosyncrasies in which the reasons for the differ- cent outcome may lie. Conversely, for most different systems with the same ‘outcome (MDSO) the white areas indicate their specific conditions, whereas the shaded areas for two or three cases respectively indicate their remaining ‘commonalities in which the reasons for the same outcome may be sought. Until recently, however, such designs had not been fully operationalized tnd, indeed, Przeworski and Teune themselves did not specify any concrete case selection procedures. It also must be noted that itis only by linking the specific outcomes to these designs that they can approximate Mill’s “quasi- experimental” methods. The MSDO design (most similar, different outcome) in this respect is mostly applicable for “very small-N" situations, where paired comparisons or the systematic matching of very few cases (often just three or four) may lead to a narrowing down of the “conditions of occurrence” for exploratory purposes, in order to identify some factors that may possibly be (COMPARATIVE RESEARCH DESIGN B soo 'MDSO Most Difefent, Similar Outcome MSDO Most Similar, Diterent Outcome Figure 2.1 Most Different and Most Similar Systems Designs responsible for the respective outcome. Conversely, the MDSO technique (most different, similar outcome) can cover an already somewhat larger but still limited number of cases in an order of magnitude of, say, 15 to 25, such as the EU member states or similar still sufficiently comparable groups of ‘cases. This design allows for the testing of hypotheses in this somewhat larger universe showing their “external validity” and the specific realm of already somewhat more developed “medium-range” theories? Further Guidance for Case Selection ‘An important difference between the case-oriented and relatively small- cr intermediate-N approach presented here and the usual large-N statistical “variable- oriented” analyses lies in the fact that case selection by itself is a process guided by the underlying research question and the preliminary hypotheses ‘one may have in this respect. This process of case selection, therefore, can be as tentative and iterative as the variable selection and model specification in statistically oriented research. Thus, in small- or intermediate-N situations the selevant population of cases cannot simply be constituted by purely mechan cal procedures like, for example, random sampling. Even if the two mi general criteria mentioned above (sufficient homogeneity of the universe of ceases considered and maximum heterogeneity within this universe) are taken into account, the inclusion of each case should be justified on theoretical4 CONFIGURATIONAL COMPARATIVE METHODS grounds (Mahoney & Goertz, 2004; Ragin, 1994). This also means that the number of cases analyzed often cannot be fixed a priori, but new cases may be added, or others dropped, in the ongoing process of research when new hypotheses arise that can be confirmed with more similar or falsified with ‘other contrary cases, Inthe inter-war project, for example, the case of Estonia was added at a later stage when the specific conditions for the survival of democracy in Finland could be contrasted with the breakdown of democracy in this otherwise very similar case (De Meur & Berg-Schlosser, 1996). ‘The procedure of case selection often meets discipline-specific constraints. For instance, in some disciplines, as in political science, the possible units at the macro level (e.g., states and their political systems, cities, policies) are mostly institutionally “given.” They then may or may not be selected for inclusion in the study. In other disciplines, such as social anthropology, for example, certain ethnic groups or smaller units like clans may be considered relevant, but their social boundaries may be more fluid, and a precise delin- eeation of the population of cases to be included in the study may be more difficult. Similarly, in sociology, again depending on the research question, varying units like families, schoo! classes, interest groups, and so on may be considered where the relevant population of cases may pose some problems, and the particular choices made must be made explicit (see also Ragin, 2004). Furthermore, there are, of course, a number of pragmatic considerations concerning the kind and overall number of cases selected. One's own famil- iarity with the cases, language skills, accessibility of sources and data, the possibility of cooperating with other case experts, and, last but not least, the availability of funding and other resources always play a role. This should ‘enable the researcher to develop certain degree of “intimacy” (Ragin, 1994) with each one of the cases under consideration. In other words, there must be sufficient “case-based knowledge” before engaging in the further technical operations of QCA. However, as important as case knowledge should be, the primary concem should still be the original research question and the subse- {quent use of theory to guide case selection.’ core “Good Practices” (1): Case Selection in SECC ccuts Ine ices) ‘© Make sure tha ll cases share enough background characteristics. + Make sure that you have avery clear definition ofthe outcome you are trying 120 “explain” across the cases. ‘+ Generally, tis best to Include both eases with a “positive” outcome and cases with a “negative” outcome, ‘ £ | i i t | COMPARATIVE RESEARCH DESIGN 2s *+ Dont ake your population (or sample) of cases a aves leave open the posbliy co include adeional eases orto remove caes ata ater sage of the research + Ifyou engage ina smal or inarmedate-N desi: When pondering on how italy set ou Gin afaga, wk yoursell whather you can ghin slant falar (empirical “intimacy") with each ese. ‘= Ifyou engage in a large-N design, make sure you gain sufficient familarcy ‘with the pes (kinds or categories) of eases. THE SEL CTION OF CONDITIONS In a similar manner, the selection of conditions must be guided by theoretical criteria, Here, however, we are confronted with the opposite embarrassment— namely, a potential abundance of conditions to be considered. Given the state of theory for most empirical questions in the social and behavioral sciences, and in particular the existence of numerous competing theories, a large umber of conditions often cannot be excluded a priori. This predicament is exacerbated if we seck to go beyond “universlizing” explanations and explore diversity (i... “variation finding”; see Tilly, 1984) oF address patterns of “multiple conjunctural eausaion” (See p. 8). Ideally the researcher should try to narrow his or her perspective to only a few “core” theories, but even then, the sheer number of competing “explanations” ofthe outcome of interest often remains too great. How should a researcher choose a smaller set of conditions if 100 many theories seem relevant? We discuss four concrete strategies. “The first conventional strategy for limiting the number of conditions is to test any relevant hypothesis for the outcome concerned in a strictly “Popperian” falsficatory manner (see p. 3). Thus, for example, the well- known thesis that “the more well-to-do a nation, the greater the chances that it will sustain democracy” (Lipset, 1960, p. 31) cam be tested in this way. In the contemporary world, this hypothesis is verified for about 70% of the cases, that are relatively weil established and consolidated democracies. However, the hypothesis does not account for poor countries with relatively stable democra- cies like Botswana, India, and Papua New Guinea or forthe failure of democ- racy in relatively well-developed ones like Weimar Germany. Indeed, in the imter-war project, a specific test of this hypothesis across Europe produced score of only 10 out of 18 “correct” results (Berg-Schlosser & De Meur, 1994), Following a second, complementary strategy, such a hypothesis must thus be ‘modified and specified more closely if tis to satisfy any epistemological erte- fia more demanding than vague “probability.” One way to do this is to test “conjunctural hypotheses” in which the selection of conditions is guided by26 CONFIGURATIONAL COMPARATIVE METHODS éxplanations that are combinatorial in nature (see, e-g., Amenta & Poulsen, 1904; see also p. 125). In this way, certain constellations of conditions— for example, concerning conditions favorable to democracy in a number of poor ‘countries or unfavorable in some richer countries—can be identified and tested. Third, widening the horizon still further, investigators may adopt a “per- spectives” approach—that is, supplying a mixed bag of conditions derived from the main theoretical perspectives in the empirical literature. This approach is probably the most common way of dealing with complex prob- lems in empirical social scientific research. The investigator takes @ thorough ook at the state of the art in any given area and then develops a specific research design that takes the wider range of these conditions into account. ‘Atthe same time, he or she develops a way to adjudicate between competing explanations and to allow for “interaction effects” among certain conditions. To stay within our previous example of empirical democratic theory, the works of Dahl (1971; 1989) and the overview by Lipset (1994) discuss a wide range of factors that are conducive to more stable forms of democracy in the modern ‘world. These reviews do not, however, consider the relative “weight” these factors or their possible interaction effects ‘But even such broad overviews may not address all possibly relevant condi- tions or all relevant interactions, Often, therefore, researchers must adopt a fourth, “comprehensive” stretegy, relying on all existing theories, hypotheses, and explanations. Even though the “all” in this formulation can never be fully satisfied, it points to the potential for complexity on the variables side of the “many variables, few cases” dilemma (see the limited diversity problem, next page). Pursuing our example ofthe inter-war project, as this project mostly mobi- lizes a political science and sociology literature, such a comprehensive approach can be structured with broad “systems” models of the Parsonsian or Eastonian kind, which potentially comprise all relevant aspects and interactions to be con sidered. The different subsystems or categories of such a mextel can be “filled” using a theoretically and historically informed listing of conditions (as inthe ‘per- spectives” approach). Still even the comprehensive approach offers no assurance that all relevant factors and interactions have been taken into account. In the inter-war project, for example, the authors departed from a relatively comprehensive “system” model in Easton's (1965) sense, by distinguishing seven major categories of conditions: (1) general historical and geopolitical background conditions; (2) socioeconomic development; (3) social cleavage structure; (4) political-cultural traditions; (5) intermediary structures, major Interest groups, and party systems; (6) institutional setup of the central politi- eal system and the administrative apparatus; and (7) external (international) factors. Within each category. several conditions were then identified, depending ‘on the state of empirical democratic theory within each category (for further (COMPARATIVE RESEARCH DESIGN 2 details, see Berg-Schlosser & Mitchell 2000; 2003). Over and above these “system-derived” conditions, two further sets of conditions were included: some conditions to tap the crisis effects concerning the social and political turbulences in many countries (formation of new states, establishment of new democratic regimes, and the effects of the Great Depression after 1929) and some specific actor-related conditions (major interventions by leading person- alities or collective actors such as the military). Obviously, it was not possible to process this very long list of conditions (61 “system-derived” conditions plus 4 crisis- and 8 actor-related conditions) in one single model for further empirical analysis, especially in a small- or intermediate-N setting In this process of condition selection, itis very important to keep the number of conditions quite low, especially in small- or intermediate-N research designs ‘The key issue here is not so much the absolute number of conditions but rather the ratio between their number and the number of cases. Consider some binary conditions (dichotomized: only “0” or “I” values). As their number increases, the number of possible combinations of these variables increases exponentially (De Meur & Rihoux, 2002; Ragin, 1987). If there are only two conditions, there are only four combinations, But then the figures rise very quickly: for instance, 8 combinations (ic., 2°) for 3 conditions, 64 combinations (i.e. 2%) for 6 condi tions, 512 combinations (1¢., 2) for 9 conditions, and so on ‘Thus the number of possible logical combinations of conditions can quickly ‘exceed the number of cases, and the empirically observed cases will occupy only a tiny proportion of the potential “logical space” (as shall also be demon- strated further with Venn diagrams; see Chapter 3, Figures 3.2 and 3.3). This is the limited diversity problem: The observed data are far less rich than the poten- tial property space delineated by the conditions. The danger, if we proceed and use these data to perform QCA treatment, is that we shall obtain an individual explanation for each individual case. In the extreme, only a description of the cases—of each separate case—will be obtained, rather than a genuine explana- tion. For this reason, it is better to select a limited number of potential condi: tions. This further reinforces the more general argument in favor of parsimony (Gee p. 10): The fewer the number of “causes” we need to explain a phenome: non of interest, the closer we come to the “core” elements of causal mecha: nisms. Moreover, the better we are able to identify fundamental causes, the ceasier it will be to produce results that may be tested on other eases, and even- tually corroborated or falsified. It is exactly this “falsifiability” of results that ives a method its scientific quality (Popper, 1963; see also p. 3). ‘The initial complexity of possible factors influencing the outcome can also be reduced by a number of stepwise multi-methodological procedures. In the inter-war project, for example, the seven major categories of conditions were tested separately with esQCA but also with other outcome-oriented techniques28 (CONFIGURATIONAL COMPARATIVE METHODS ‘COMPARATIVE RESEARCH DESIGN » such as Discriminant Analysis (a statistical technique) to identify strong 7 country, several specific problems must be addressed. These follow from thd bivariate relationships. In a second step, only conditions displaying strong necessity of measuring the proximity or remoteness of pairs of cases in th rebionships were retzined across all eategorics and tested again, lao show heterogeneous, multidimensional space defined by the conditions. These dist ing some specific combinations of factors. Other procedures, lke confirmatory tance measures provide the basis for determining the “most different” and | factor analysis, may also be employed to process together conditions that load “most similar” pairs or groups of cases with regard to the outcome. The two on the same dimension—for example, values of urbanization, indusriatiza- main issues are (a) choosing from among a variety of different ways of mes. tion, and literacy being combined to a single measure of “modernization.” suring the distance between pairs of cases in a multidimensional space and Some other factors could also be combined by logical procedures, such as (©) assigning relative weights to the conditions that define this space. In this taking together the existence of large landlords and a rural proletariat as @ ‘way the complexity ofthe data can be retained as much as possible in the com common factor of feudal or quasi-feudal patterns of landholding. In this way, plexity of the proximity measure. the 61 original conditions were reduced to 8 “super conditions,” also corre- ‘The purpose of this section is to present, in a nutshell, a systematic proce sponding to major tenets in empirical democratic theory, and the relationship dure to achieve these goals—it is called MSDO/MDSO. Ultimately, it also ‘between the number of conditions and the number of cases considered thus, allows, through the systematic matching and pairing of eases, to identify soms became much more favorable (De Meur & Berg-Sehlosser, 1996). ey conditions that can either be used for further, more qualitative, interpret tion of some clusters of cases or be considered as @ way to identify “core conditions that can be used in a subsequent application of QCA. Thy cele | __ MSDO/MDSO procedure is designed for research situations in which the con Serre an Cae) Re eon ditions (“explanatory factors") are—or can be—grouped into categories or Rear ee eee and clusters, Its also particularly helpful when conditions are numerous, H Do not include a condition that does not vary across the cases. In other words,"a variable must vary" otherwise It isa constant. Keep the number of conditions relaively low. A large number of conditions tends to “indvidualize’” each case, making It dificult to find any regularity or Cer] TRA oa Soak any synthedie explanation of the outcome across the cases. ‘Altogether, a good balance must be reached between the number of cases and the number of conditions. The ideal balance is not a purely numerical fone and will most of the time be found by trial and error. A common practice, in an intermediate-N analysis (say, 10 to 40 cases) would be 10 select from 4 to 6-7 conditions For each condition, formulate a clear hypothesis regarding its connection to the outcome possible, formulate this hypothesis in the form ofa statement. about necessity andlor sufficiency MSDO/MDSO: A SYSTEMATIC PROCEDURE FOR MATCHING CASES AND CONDITIONS ‘To establish which cases are “most similar” or “most different” in any given research situation, rather than proceeding from merely intuitive “hunches” or purely pragmatic reasons like knowing the history and language of a particular Full deta with the “interwar project” example (with seven clusters of con- ditions), can be found via the resources Web page (see the Appendix). See also De Meur and Berg-Schlosser (1994) and De Meur (1996) on this same example, as well as De Meur, Bursens, and Gottcheiner (2006) for another ‘empirical application. 1. Preparing the data: Each variable (conditions and outcome) is dichotomized, for each cate. 2. Computing “distance matrices” between pairs of countries, for exch cluster fof conditions. As a measure of distance, a "Boolean distance” is used: the number of Boolean (\e. dichotomized, with 0 or 1 values) conditions by | “which two cases differ from one another. This makes It possible to identify, for each of the seven distance matrices, che minimum distance obtained for ‘countries with diferent outcomes (MSDO) and the maximum distance for: ‘countries with the same outcome (MOSO)—some key pairs of MSDO cases and MDSO eases are thus singled out (Continued)0 CONFIGURATIONAL COMPARATIVE METHODS. (Continued) 3. Aggregating the dita natries The results ofthe previous step are agure- gated, considering all sven clusters of condiions—this produces a compre- hensive tance matric 4. Defining levels of (dissimilar: The comprshensive. distance: matrix is marked at different levels. First, for each pair of countries the clusters of Conditions in which the distance is equal to either the minimum (for Councries with a different outcome) orto the maximum {for countries with the same outcome) ara identified. These are the “level 0” (i.e. strongest. “purest?) similarities or dlesimiartes. Lower levels of slaiarcy (or disim\- larg) can also be taken into account (“evel 1 “level 2” etc). Thus, difer- cent levels of (as)simiarty can be obtained forall dusters of conditions Enabling the computation of lower levels of (dssimirey allows one to main- ‘ain a broader “reserve” of conditions to be kept for the next steps. 5. Synthesizing (dis)similarcy:The information across ll seven clusters of cases is synthesized so as to obtain a complete pisurs of the (dis)similrities ofall the cases, within the MOSO zone for cases with the same outcome, and ‘within the MSDO zone for cases with a different outcome, 6 Producing overall similarity and dissimilarity graphs: For each level of (ds)simiary (cee Step 4), only thse pis of cases that are mest (d)si- lar Ge. that dsply the largest number of cers of conditions with hgh- ‘xt levels of (is) tae) are retained This formation s then transated inco a "simlanty graph” or a *dssimlargy graph” which vivazes the Constelitions of cases with che greater (ds)imiaries. Those graphs are tventualy merged nto aggregated graphs in which the diferent levels of (dis)similarity are superimposed. 7. Systematic matching and contrasting f cases: the basis ofthe established (Gis)simarves and similares, a final tap. the most differen eases with the same outcome (MOSO) and the most simihr cases witha diferent out- come (MSDO) can be selected is chen possble to lst the individual con- ditions (n exch cluster of conditions) that characterize the remaining (dis)similares. I is among these selected condions thatthe reason for the common (or contrasted) outzome may be determined. ‘A practical note:At this stage of development, te technical steps and termi nology ofthis procedure are not s0 easy to grasp for users not specifically ‘walned in formal language: They areas qui iberncensive i they haye tobe ddone manu This is wlysofeware for automating these procedures has been developed (see resources Web page listed in the Append). (COMPARATIVE RESEARCH DESIGN a Once this stepwise procedure has been completed, it is then possible to reexamine specific groups of cases (MDSO or MSDO cases), pairwise, three by three, and so on, and to use qualitative judgment and “thick” case knowl- edge to examine the remaining (dis)similarities in geeater detail. Thus, the final step after the MSDO/MDSO procedure can be, on the one hand, a re-interrogation of some key groups of cases (of the MDSO or MSDO type, depending on the goals of the researcher) in a more qualitative way and also in a more focused way, because what needs to be interpreted is specifically the interplay of some more crucial conditions that have been singled out through the MSDO/MDSO procedure. Such more qualitative, historically informed interpretation can yield some “causal” insights as to how and why those core conditions have led (or have not led) to the outcome of interest in the contrasted or similar cases. On the other hand, the MSDO/MDSO procedure can also be used as a prior step, in the process of model specification, before one engages in QCA proger. Indeed, MSDO/MDSO identifies some particularly crucial conditions that seem to be at the heart of what brings together (or differentiates) cases with a similar (or different) outcome. For instance, using the inter-war project data, selections of such conditions could be used for various QCA tests, To sum up, the MSDO/MDSO procedure provides a futher extension, ‘operationalization, and application of Mill’s methods, including the possitil- ity of “conjunctural” patterns of causation and the implications for the neces- sary “dialogue between theory and data” in a “‘case-oriented” manner, It can be used to produce substantial tests and further refinements of theory when ‘many competing theories abound (see also Berg-Schlosser, 2004). It leaves some room for flexibility—for example, allowing for more or less stringent criteria of (dis)similarity, depending on how restrictive the researcher wants to be with regard to the number of conditions to be Kept for further analyses. ‘+ From the outset, sufficiently homogeneous domain of investigation ‘must be delineated, from which cases wil be selected. ‘+ A clear defiition ofthe outcome of interest must be lad out at avery ‘arly stage of che analysis, because itis indispensable for case selection. * In most small- or intermediate-N research designs itis useful to have ‘meaningful variation in the outcome—that is,to select cases with both “negative” and “positive ourcomes.2 ‘CONFIGURATIONAL COMPARATIVE METHODS + ts important to keep the number of conditions reasonably low; the ‘greater the number of conditions, the more one runs the risk of “individualzing” the explanation of each case. 1 There are diferent, complementary strategies for reducing the number of conditions when there are many “candicate” theories. ‘+ The MSDOMDSO procedure (systematic matching and contrasting of cases) can be used either for further, more qualitative, incerpretation of clusters of eases or asa prior step to further QCA analyses—by Iderdfying crucial conditions. ‘= MSDOIMDSO is best suited for research situations in which the conditions ean be grouped into clusters. ‘+ MSDOIMDSO leaves room for flexibility in particular with regard to the stringency of criteria of (ds)similarity across cases. Key Complementary Readings i Berg.Schlosser & De Meur (1997), De Meur & Bery-Schlosser (1994), De Meur et al. (2006), Ebbinghaus (2005), Mahoney & Goerez (2004), Praeworski & Teune (1970). NOTES 1. Note that “units” do not always equate with “cases.” For instan ‘anaiysis) could be & country, whereas a unit of observation within that case could be & tiven region or a given political institution, or that country ata given point in ime. 2. A somewhat different aspect of “most different systems designs” when they are applied to “multi-level analyses"—linking, for example, the "meso" level of party systems 10 the “macro” overall country level—is discussed by Tiemann (2003). 5. For futher elaborations with regard to case selection in small and intermedite-N stings, see Ebbinghaus (2008), Gerring (2006), Mahoney and Goertz (200). 44. Note that this advice to keep models sufficiently “short” is also valid in statist cal research, For example, Achen (2005) argues that remeih ‘analysts should work with five or six independent variables case (unit of Crisp-Set Qualitative Comparative Analysis (esQCA) Benoft Rihoux Giséle De Meur emake ‘After reading this chapter, you should be able to: "© Understand the key operations of Boolean algebra and ute the correct Conventions ofthat language ‘+ Transform tabular data into Venn diagrams and vice vers; interpret Venn diagrams ‘+ Replicate standard «QCA procedure, step by step, using the sofware ‘+ In the course of this procedure, dchotomize your variables (conditions and outcome) in an informed way.and ata later stage, use an appropriate strategy for solving “contradictory configurations” Weigh the pros and cons of using logical remainders (non-observed cases) ‘= Read and interpret the minimal formulas obtained at the end of the Qc . esQCA was the first QCA technique developed, in the late 1980s, by Charles Ragin and programmer Kriss Drass. Ragin’s research in the field of historical sociology led him to search for tools for the treatment of complex sets of binary data that did not exist in the mainstream statistics literature. He adapted for his own research, with the help of Drass, Boolean algorithms that hhad been developed in the 1950s by electrical engineers to simplify switching circuits, most notably Quine (1952) and McCluskey (1966). In these so-called minimization algorithms (see Box 3.2), he had found an instrument for identi- fying patterns of multiple conjunctural causation and a tool to “simplify com. plex data structures in a logical and holistic manner” (Ragin, 1987, p. viii. BM CONFIGURATIONAL COMPARATIVE METHODS esQCA is the most widely used QCA technique so far. In this chapter, a few basic operations of Boolean algebra will first be explicated, so the reader can ‘grasp the nuts and bolts of esQCA. Then, using a few variables from the inter war project (see Chapter 2), the successive steps, arbitrations, and "good prac tices” of a standard application of esQCA will be presented. ‘THE FOUNDATION OF CRISP-SET QCA: BOOLEAN ALGEBRA IN A NUTSHELL jorge Boole, a 19th-century British mathematician and logician, was the fist to develop an algebra suitable for variables with only two possible values, such as propositions that are either true or false (Boole, 1847; 1958 (1854)) Following intuitions by Leibniz one century before him, Boole is the origina- tor of mathematical logic that allows us to “substitute for verbal reasoning a genuine symbolic calculation” (Diagne, 1989, p. 8). This algebra has been Suclied further by many mathematicians and logicians over the last few decades, Iehas been central tothe development of electronic circuits, computer science, and computer engineering, which are based on a binary language, and has led to many applications, mostly in experimental and applied scientific disci- plines. Only a few basic principles and operations will be presented here (for more details, see, ¢.g,, Caramani, 2008; De Meur & Rihoux, 2002; Ragin, 1987, pp. 89-123; Schneider & Wagemann, 2007, forthcoming). Boolean algebra, a5 any language, uses some Conventions that need to be understood before engaging in esQCA. Cor Neo eIeIeeuc Recut 41. The main conventions of Boolean algebra are as follows: ++ An uppercase leter represents the [1] value for a given binary variable. “Thus [A] is read as: "variable A is large, present, igh, 2" ‘= A lowercase leer represents the {0] value for a given binary variable. “Thus (a is read as:"variable A is small absent, low, + A dosh symbol [represents the “don's cara” value for a given binary ‘variable, meaning it ean be either present (1) or absent (0) This also ‘could be a value we don't know about (eg, because i is irrelevant or ‘the data is missing). Ie not an incermediats value becween [1] and [0]. (CRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) » 2. Boolean algebra uses a few basic operators, the ewo chief ones being the following: ‘+ LogicalAND represented by the [*] (mukplication) symbol. NB: Ie can also be represented with the absence ofa space: [A*B] can also be wrk tem as: [AB]. + torial "OR, represented by the [4] (ado) symbol 3, The connection becween conditions and the outcome: The arrow symbol [41 is used to express che (usually causal link besween a set of conditions fon the one hand and the outcome we are trying to “explain” on the other, With this very basic language, itis possible to construct very long and elab- orate expressions and also to conduct a complex set of operations. One key ‘operation, which lies at the heart of esQCA, is called Boolean minimization Box 3.2 Mira ce een leis the “reduction” ofa long, complex expression into a shorter, more pars ‘monious expression It can be summarized verbally as follows:"if two Boolean ‘expressions differ in only one causal condition yet produce the same outcome, ‘then the causal condition that distinguishes the two expressions can be con- sidered Irrelevant and can be removed to ereate a simpler, combined expres- sion” (Ragin, 1987, p. 93). Let us use a very simple example. Consider the following Boolean expression, with three condition variables (R, Band I) and ‘one outcome variable (O) (Formule 1): R*B*l + REBT > O ‘ ‘This expression can be read as follows:"[The presence of R, combined with the presence of B and with the presence off] OR [The presence of R, combined ‘with the presence of B and wich the absence of I lead to the presence of out come O" ‘Notice that. no mater which vole the [] conditan takes (0 or 1), the [I] out- ‘come value isthe same. This means, in verbal reasoning, that the [l] condition is superfluous:ic can thus be removed from the intial expression. Indeed, if we remove the [] condition, we are left with a much shorter reduced expression (whichis called a prime implicant) (Formula Z) R*B30 (Continued)% CONFIGURATIONAL COMPARATIVE METHODS (Continued) ic is read as follows: “The presence of R, combined with the presence of B, leads to the presence of outcome ©” This reduced expression meets the parsimony principle (6 p. 10). We have been able to explain the O phenome- ron in a more parsimonious way, but stil leaving room for complexity, because for O to be present.a combinaton of the presence of Rand the presence of B In other words, to relate this again to necessity anid suiciency (se p. 10) the presence of Ris necessary (but not suficient) for the outcome: tkewise, ‘the presence of B Is necessary (but not sufcent) for the outcome. Because neither ofthe twa conditions is sufficient for the outcome they must be com bined (oF “intersected” though Boolean multiplication, see Box 3.1) and, together they could possibly? form a necessary ond sufficient combination cf condition leading to the outcome. Boolean minimization ean also be grasped in a more visual way. Let us ton- sider again the same example and provide more explicit labels for the three conditions: respectively, R stands for “RIGHT,” B stands for “BELOW,” dnd I stands for “INSIDE.” As these are binary conditions, each divides the universe of cases into two parts: those that meet the condition (value 1) and those that do not (value 0). So we have three conditions: ‘© RIGHT (1 value) versus notright (0 value) ‘+ BELOW (I value) versus not-below (0 value) 1 ‘© INSIDE (1 value) versus notinside (0 value) | Because each condition divides the universe into two pars, these of three conditions divides it in2 * 2" 2= 8 zones, called “elementary zones” Within tach zone, the values for conditions of each case are the same. Let us also Suppose that we know the value ofthe outcome variable fo each one of these 8 zones. The data can first be presente in tabular format (see Table 3.1) ‘The first column of Table 3.1 contains the case labels (“caseid”), from easel" to“case8” The following three columns contain all logically possible combinations (here are 8 of them—ie, 2% se p. 27) of the binary RIGHT, BELOW, and INSIDE conditions. Finally, the fifth coh contains the value ofthe OUTCOME foreach on ofthe 8 combinations, The fst sx cases dis- play a [0] outcome (Boolean notation: {outcome while the last owo display 4 [1] outcome (Boolean notation: [OUTCOME]). The Boolean expression that teas minimized above (see Formulas I and 2in Box 32.) eoresponds simply to a anslation of the two botiom rows of this table “erapmscmrersoaunna sy apte ew (CRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) a Table 341 Rave Data Table 3-Condition Example) caseld Right Below Inside Outcome case 0 ° ° 0 cased 1 0 ° 0 cased 0 0 1 ° eases 1 0 1 0 eases 0 q ° ° cas06 0 1 1 ° ease 1 1 1 1 cased 1 1 ° 1 This same data can be represented in a visual way, in a Venn diagram (see also p. 23), showing visually that each condition cus the universe of eases in two (Figure 3.1) and making the 8 basic zones visible. In this Venn diagram: ‘+ Condition {RIGHT cuts the space vertically: All cases on the right-hand side of the vertical line have a (1) value for this condition, whereas all cases on the lft- hhand side of this line (“not right”) have a (0) value for this condition. ‘+ Condition [BELOW] cus the space horizontally: All cases below the horizontal line have a (1) value for this condition, whereas all cases above this line (“not below”) have a (0) value for this condition. ‘+ Finally, condition [INSIDE] cus the space between what's inside or outside the Square in the middle: All cases inside the square have a (1) value for this condi- tion, whereas all cases outside the square (“not inside”) have a (0) value for this condition, This Venn diagram also contains information about where each case is located. In this simple example, cach basic zone is occupied by a single case. In addition, the diagram contains information about the value of the outcome variable, Let us consider only the [1] outcome. It corresponds to the dark- shaded area, at the bottom right of the diagram, and corresponds indeed toEy CONFIGURATIONAL COMPARATIVE METHODS = Figure 3.1 Venn Diagram Corresponding to Table 3.1 (3 Condition Variables)* * Venn diagram produced bythe “visualize” tool, TOSMANA 1.3.0.0 software ccase7 and case8 (as in the last wo lines of Table 3.1). This light shaded area ccan be expressed, in Boolean notation, as follows (Formula 3) RIGHT * BELOW * INSIDE + RIGHT * BELOW * inside > OUTCOME [Note that this formula is exactly the same as Formula 1. Itis read as follows: “The outcome is present for case7 [a ease that is on the right AND below AND inside} OR for case8 [a case that is on the right AND below AND not inside].” ‘This is @ long formulation, which requires the description of each basic zone: It describes first the basic zone where case7 stands, then the one where case8 stands, Technically speaking, this formula contains two rerms (each term, is a combination of conditions linked by the “AND” operator), and each term contains all three conditions. This is precisely where the Boolean minimiza- tion intervenes: It makes possible describing these two zones with one simpler ‘and shorter expression, consisting of only one term. Indeed, we notice, visu ally, thatthe two basic Zones where cases 7 and 8 stand, when combined, form CCRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) x» 4 Jarger square: All the cases that are located below the horizontal line, AND all the cases that are located on the right-hand side of the vertical line. This larger zone can be expressed as (Formula 4): RIGHT * BELOW > OUTCOME Note that this formula is exactly the same as Formula 2. It means that we simply need to know about two conditions—[RIGHT] and (BELOW}—to account for the outcome, which corresponds to cases 7 and 8, We need not ‘know whether or not those cases are inside or “not inside” the middle square, because this information is superfluous: What is shared by these two cases ‘with a [1] outcome is that they are “on the right AND below.” So we ean simply remove the [INSIDE/not inside] condition. [As we shall sce later, this is exactly the operation that is performed by the software on more complex data sets, with more conditions, and also with “empty” basic zones (i.e, with no cases observed). Of course, this makes the ‘operations somewhat more complex, so itis best to let the computer perform the algorithms (in esQCA, the software uses the Quine minimization algorithm), Now that we have introduced the basies of Boolean language and opera- tions, we ean present key practical steps of esQCA. We shall pursue the same example as in Chapter 2, from the inter-war project, where the cases are 18 European countries. STEP 1: BUILDING A DICHOTOMOUS DATA TABLE, Of course, building a relevant data table requires previous work: A well- thought-out comparative research design, and im particular rigorous case and variable selection (sce Chapter 2). Remember, also, that at this stage the researcher is supposed to have gained adequate substantive knowledge about each case and theoretical knowledge about the most relevant variables (condi- tions, in particular) included in the analysis. Among the large variety of approaches dealing with the more general conditions favoring the emergence and consolidation of democratic political systems in different parts of the world (see p. 26: list of main categories of con- ditions in the inter-war project), we have selected one that addresses overall socioeconomic and “structural” factors. Of course, for a more comprehensive account other factors such as specific historical and cultural conditions, inter- ‘mediate organizations, institutional arrangements, actor-related aspects, urd $0 ‘on must also be considered (for an application of such a more comprehensive0 CONFIGURATIONAL COMPARATIVE METHODS design, see also Berg-Schlosser, 1998). For purposes of illustration, however, using the selected approach will suffice—in shor, for the sake of clarity, we need a relatively simplified theory, which does not entail too many conditions. "As a reminder, the specific outcome we are trying to “explain” here isthe sur- vival or breakdown of democratic systems in Europe during the inter-war period. Why is it that some democratic systems survived, while others collapsed? In QCA terminology, this variable we seek to explain is called the outcome. ‘The most influential study dealing with the more general socioeconomic preconditions of democracy was S. M. Lipset’s Polirical Man (1960), in particular his chapter “Economic Development and Democracy.” There, he (re)stated the general hypothesis that “the more well-to-do a nation, the greater the chances that it will sustain democracy” (p. 31). Indeed, among the “stable European democracies” analyzed by Lipset were cases like Belgium, the Netherlands, Sweden, and Great Britain, which all showed high levels of ‘wealth, industealization, education, and urbanization. Under his (very broad) category of “unstable democracies and dictatorships" figured countries like Greece, Hungary, Italy, Poland, Portugal, and Spain, with lower levels in this regard. However, he also noted that Germany issn example ofa nation where growing industialization, urbanization, ‘wealth and education favoured the establishment of a democratic system, but in ‘which a series of adverse historical events prevented democracy from securing legitimacy and thus weakened its ability to withstand criss. (p. 20) ‘This statement certainly applies to Austria as well, but the kind of “adverse historical events” and their specific roots were not investigated by Lipset. Similarly, the fact that countries like Czechoslovakia, Finland, and France (which also had higher levels of development and democratic institutions, and ‘which, as far as internal factors were concerned, survived the economic crisis of the 1930s) were grouped by Lipset in the same “unstable” category, was not very helpful from an analytical viewpoint. In later years, Lipset's work was followed by a number of conceptually and statistically more refined studies ‘and drew considerable criticism as well. However, when he later reviewed his ‘original study, he still found its basic tenets confirmed (Lipset, 1994; see also Diamond, 1992). For each of the four main dimensions discussed by Lipset (wealth, indus- trialization, education, and urbanization), we have selected one major indicator as listed in Table 3.2. and have provided the data for each one of the 18 cases (countries) considered. In this example, we have 18 cases (each case being a row in Table 3.2) ‘The outcome variable (SURVIVAL) is already of a dichotomous nature: (CRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) a Table 32 _Lipset’s Indicators, Raw Data (4 Condi casei | Gnpcar | ursanrza | Lireracy | INDLAB_| SURVIVAL AUS 70 3a 98 334 ° BEL 1098 05 344 489 1 Cz. 386 o 959) 374 i EST, 468 285, 95 14 o IN 590 2 99.1 2 1 FRA’ 983 212 962, 348 A GER 795 565 98 04 ° GRE 390 311 592 28.1 ° HUN a4 363 85 216 ° TRE (662 2 95 145 1 TTA 517 314 a 296 0 NET. 1008) 788 99 393 L POL 350 37 769) 2. 0 POR 320 153, 38 Ba 0 ROM 331 219 618 22. o ‘SPA 367 a 556 255 oO SWE 897 En 999) 323 1_| UK 1038 7” 99 99) 1 Labels for conditions * CCASEID: Case identification (country name) abbreviations: AUS Ausra; BEL Belgium: CZE CCuechoslovaka; EST Estonia; FIN Finland; FRA France; GER Germany: GRE Greece; HUN iangary; IRE Wieland; ITA Haly; NET Netberlands: POL Poland; POR Portugal; ROM Romania ‘SPA Spain SWE Sweden; UK United Kingdom GNPCAP: Gross National ProductCopita (ca. 1930) LURBANIZA: Urbanization (population in towns with 20,000 snd more inhabitants) LITERACY: Literacy INDLAB: Industrial Labor Fore (inlading mining) (Detaits of sources in Berg Schloser& Mitchell, 200, 2003)2 ‘CONFIGURATIONAL COMPARATIVE METHODS The (0] outcome value stands for “breakdown of democracy” (For 10 cases), and the [1] outcome value stands for “survival of democracy” (for the other 8 cases), However, the four variables that are supposed to “explain” the outcome, Which are called conditions’ in QCA terminology, are continuous (interval- level) variables. To be used in esQCA, those original conditions must be dichotomized according to relevant thresholds. To dichotomize conditions, itis best to use empirical (case-based) and the coretical knowledge (see “good practices” box, below). In this example, we have chosen to set the dichotomization thresholds as follows + (GNPCAP}: Gross National Product/Capita (ca. 1930): 0 if below 600 USD; 1 if above. * [URBANIZA]; Urbanization (population in towns with 20,000 and more inhab- itants): 0 if below SOE; 1if above. + [LITERACY]: 0 if below 75%; 1 if above ‘© (INDLAB}: Industrial Labor Force (inel. mining): 0 if below 30% of active population; 1 if above, Cee "Good Practices” (3): How to Dichotomize feet eee eet ‘Always be transparent when justifying thresholds. les best co justify the threshold on substantive andior theoretical grounds. I his is not possible, use technica! criteria (e.g,,considering the distribution ff cases along a continuum; see also p. 79). As a last resort, some more ‘mechanical cutoff points such as the mean or median can be used, but one should check whether this makes sense considering the distribution of the ‘Avoid artificial cuts dividing cases with very similar values. More elaborate technics! ways can also be used, such as. clustering techniques (see p. 130), but then you should evaluate to what extent the ‘lustars make theoretical or empirical sense, [No matter which tachnique or reasoning you ute to dichotomize the ‘conditions, make sure to code the conditions in the correct "direction," ‘thae their presence ([1] value) is theoretically expectad to be associated with a positive outcome ((1] outcome value). { CCRISP-SET QUALITATIVE COMPARATIVE ANALYSIS ( s0C) 6 We thus obtain a dichotomized data table (Table 3.3). Already before ‘engaging in esQCA proper, exploring this table visually, in a non-formal way, we can easily see that some cases corroborate very neatly the Lipset theory, by looking at the most extreme cases ‘Table 33 Lipset’ Indicators, Dichoromized Data (4 Conditions) CASEID. GNPCAP | URBANIZA | LITERACY | INDLAB ] SURVIVAL AUS - 1 0 1 L 0 BEL | L 1 1 1 1 ze ° 1 1 1 1 EST 0 0 1 0 o 7 FIN 0 o 1 oO i FRA 1 ° 1 1 [oa GER 1 1 1 | 1 0 GRE ° 0 o ° ° HUN ° 0 1 ° 0 IRE 1 o | 1 ° 1 TA ° ° 0 0 o NET 1 1 1 1 1 POL 0 ° 1 o o POR 0 0 o 0 0 | ROM o ° 0 ° 0 SPA © ° ° 0 0 SWE 1 0 Ties 1 [ux 1 1 1 1 1“ CONFIGURATIONAL COMPARATIVE METHODS For instance, Belgium (row 2) is a “perfect survivor case,” corroborating Lipset’s theory: A [1] value on all four conditions leads to the [1] outcome value [SURVIVAL]. Conversely, Portugal (row 14) is a “perfect breakdown cease,” corroborating Lipset's theory: A (0] value on all four conditions leads to the [0] outcome value [survival]. However, for many other cases, the picture looks more complex. STEP 2: CONSTRUCTING A “TRUTH TABLE” ‘The first step for which we need csQCA proper (in terms of specific software treatment®) corresponds to a first “synthesis” of the raw data table. The result thereof is called a truth table. tis a table of configurations—remember that & configuration is, simply, a given combination of conditions associated with a given outcome. There are five types of configurations, each of which may cor- respond to none, one, or more than one case. Cer] Five Types of Configurations ‘Configurations with a [1] outeome (among the observed cases}also called 1 configurations” Configurations with a [0] outcome (among the observed cases} also called 0 configurations” Configurations with a “-" ("don't care") outcome (among the observed ‘ases): also called “don't care configurations” Ie means thatthe outcome is indeterminate. This isto be avoided, since the researcher Is supposed to be ineerested in expl ing a specific outcome across wall-selected cases?” Configurations with a « C » (« contradiction ») outcome, called con- tradictory configurations. Such a configuration leads to a « 0 » outcome for some observed cates, but £0 a « 1.» outcome for other observed cases. ‘This is a logis! contradiction, which must be resolved (see p. 48) before ‘engaging further in the csQCA. Finally.configuratons with an“L” or “R” (“logical remainder’) outcome. There are logically possible combinations of conditions. that have not been ‘observed among the empirical eases. ‘Table 3.4 displays the truth table corresponding to the dichotomized data in Table 3.3 (CRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) Table 34 Truth Table of the Boolean Configurations 6 T T caseio | aNpcap | urpaniza | Lrreracy | INDLAB_| SURVIVAL SWE, 1 o 1 1 c FRA, AUS FIN,HUN,| 0 POL, ES} BEL, NET, UK, GER, CZE 0 1 L 1 1 TTA,ROM,| 0 0 0 ° ° POR, SPA, cake a|| IRE 1 0 1 ° 1 Pere “Good Practices” (4): Things to Check to Assess the Quality of a Truth Table 1 Check again that there is a mi of cases wth a“postve” outcome and cases ‘with a"negativ outcome (see Box 22, good practices forthe cate selection). “= Check that there are no counterintuitive configurations In this example, these would be configurations in which all [0] tondltion values lead to 1] outcome, or all [1] condition values lead to 2 [0] cutcome + Check for eross-condition diversity; in particular, make sure that some nditions do not deplay exactly the same values across all cases f they ‘do, ask yourself whether those conditions are too “proximate” to one another (if hey are, chey can be merged). "= Check that chere is enough variation for each condition (a general rule atleast 1/3 ofeach value) (see also Box 23,go0d practices for condition selection" variable must vary". Hone of these criteria is not met, reconsider your selection of cases andlor ‘conditions or possibly the way you have defined and operationalized the outcome. te Is also useful at this stage, to check for the necessity and sufficiency of ‘each condition with regard to the outcome.4 CONFIGURATIONAL COMPARATIVE METHODS ‘This truth table (Table 3.4) shows only the configurations corresponding to the 18 observed cases. It already allows us to “synthesize” the evidence sub- stantially, by transforming the 18 cases into 6 configurations. We find out the following: + "There are two distinct configurations with a [1 outcome, corresponding respec- tively to Czechoslovakia and Irland. ‘There is one configuration with 8 [0] outcome, corresponding to five eases (Ital Romania, Portugal, Spain, and Greece), I fits quite neatly with the Lipset theory, because a [0] value for all four conditions leads to a 0] outcome (breakdown of democracy) We also notice that there are 3 contradictory configurations (the first 3 rows in the truth table), corresponding to no less than 11 cases out of the 18. In other ‘words: Lipset’s theory—in the way we have operationalized it, at least—does not enable us to account for 11 out of 18 cases. The third contradictory configuration is particularly troubling: It contains [1] values on all of the con~ ditions and yet produces the [0] outcome for one case (namely, Germany), \whereas it produces the expected [1] outcome for the other 3 cases (Belgium, Great Britain, and the Netherlands) ‘The data in this truth table can once again be visualized through a Venn dia- ‘gram, a bit more complex than Figure 3.1 because it contains 4 conditions instead of 3 (Figure ‘This Venn diagram has 16 basic zones (configurations) —that is, 2* zones. It is constructed using the same logic as Figure 3.1. In this empirically grounded example, we can observe four types of configurations: ‘© Two configurations with a [1] outcome, covering respectively the cases of Czechoslovakia and Ireland. + One configuration with a [0] outcome, covering the five cases of Italy, Romani Portugal, Spain, and Greece ‘© Thee contradictory configurations, covering in all 11 cases (the shaded zones corresponding to the "C” label). + Finally, many non-observed, “logical remainder” (“R") configurations —10 ato- gether. Thus, there is limited diversity (see p. 27) in the data: As the 18 observed ‘eases correspond to only 6 configurations, the emmaining Boolean property space is devoid of cases. As will be shown at a later stage, these “logical remainder” ‘configurations will constitute a useful resource for further analyses. (One way to look at this evidence, from a purely numerical perspective, would be to state that the model “fits” 7 out of 18 eases. This is, however, nota correct CCRISE-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) a 0000] FRE, ITA, POR, ROM, SPA ] coor 7001 “INOLAB| | Jz 3 g (sal ua 8 ze = gure 32 Venn Diagam Conespnding Tube 34 @ Coton * Venn diagram produced by the “visualizer ool, TOSMANA 13.00 software way to proceed; remember that QCA is a case-oriented method (sce p. 6), and that each case matters. From this perspective, itis a problem that so many con- tradictions occur. Hence these contradictions first have to be resolved before pro- ceeding to the core of esQCA—siamely, Boolean minimization, AL this stage of the analysis, itis also useful to check for the necessity and sufficiency of each condition with regard to the outcome. Let us assume a ‘model that contains three conditions, A, B and C. For condition A, for instance, assessing its consistency as a necessary condition means answering the following question: “To what extent is the statement ‘condition A is nec- essary for the outcome’ consistent?” Technically, this can be computed as follows: {the number of cases with [1] value on the condition AND a [1] out. come value, divided by the total number of eases with a 1] outcome value For more details, see Goertz (2006a), Ragin (2006), and Schneider and ‘Wagemann (2007, forthcoming).8 CONFIGURATIONAL COMPARATIVE METHODS STEP 3; RESOLVING CONTRADICTORY CONFIGURATIONS It's perfectly normal to detect contradictory configurations in the course of a esQCA. It does not mean that the researcher has failed. Quite the contrary, contradictions tell us something about the cases: we are studying. By seeking ‘resolution of these contradictions, the researcher will get a more thorough knowledge of the cases (through his or her “dialogue with the cases”), be forced to consider again his or her theoretical perspectives, and, eventually, obtain more coherent data, Remember that QCA techniques are best used in an iterative way (see p. 14), Thus addressing contradictions is simply part of this iterative process of “dialogue between ideas and evidence” (Ragin, 1987). Insofar as possible, all such contradictions should be resolved or, atleast, ‘one should strive to reduce contradictions as much as possible (Ragin, Berg. Schlosser, & De Meur, 1996, p. 758)—because, eventually, the cases involved in those contradictory configurations will be excluded* from the Once again, this is problematic given the case-oriented nature of QCA. Cery ‘Good Practices” (5): How to Resolve Serr esa aeons “There are basicaly eight strategies. In realiife research, tis advisable to atleast consider all chose strategies, and most often fe will turn out that some combi- nation is useful, +1. Probably the easiest one: Simply add sonie condition(s) to the model. Indeed, ‘the more complex the model—the more numerous the condions—the less likely contradietions will occur because each condition added constiates a potential addtional source of differentiation between the cases. Of course, ‘sucha strategy should not be pursued ina “hope-and:-poke” waysit should be cautious and theoretally justified. eis advisable to add conditions one by ‘one, not to obtain £00 complex a model, Otherwise, you run the risk of ere- ating a greater problem of "lited diversity” (see p. 27) and thus of “indi “iduazing” explanations ofeach particular case; this means cae esQCA wil have mised its purpose of reaching some degree of parsimony (see p 10). 2. Remove one or more condition(s) from the model and replace i/them by (another condition(s). CCRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) ” 3, Reoamine the way in which the various conditions Included in che model are operationalized. For instance maybe thatthe threshold of dichotomization for a sven condition isthe source ofthe contradiction beeween two cates. By adjusting the threshold may be possible 20 resolve the contradiction. Akernatively, the contradiction could be due to data quality probloms-—hn that case, one could collect complementary or reved drs This che most nborintensive option but very much: te advocatid fom a Cafe-ranted perspective. 4, Reconsider the outcome variable itzelf-This strategy is often overlooked. lf the outcome hat been defined too broadly, i Is quite logical that con- tradictions may occur. For instance, Rihowx (2001) noticed, during some exploratory csQCA analyses, that his initial outcome varlable—major ‘organizational change ina given politcal parsy—could in fact be decom- posed Inco two opposed subtypes: organizational adaptation and organi- ational radicalization. By focusing the outcome solely on organizational adaptation, he Was able to resolve many contradictory configurations. 5. Reexamine.ins more qaiv and "hick" way. th ease involved in each specie contrdicory congrstion What har been missed? Whit ould dierent chose caves that hasnt been corsidered, tern he model oF inthe way the condtions or the ovcome hve been operationalized? 6. Recotsider whether al cases are indeed part ofthe same population (c.case selection, p. 20) For instance, Is a “borderline” case that is creating the contradiction, perhaps this case should be excluded from the analysis. 7, Retode all contradictory configurations as [0] on the outcome value-Ths solution, suggested by Ragin (1987), treats contradictory configurations as unclear” and hus decides to accept fewer minimizable configurations In exchange for more consistency In the cases/outcome relationship. 8, Use frequency crieeria to “orientate” the outcome, Let us consider a ‘contradictory configuration that involves ine cases. If, say it leads to a [1] ouecome for eight cases and to a [0] outcome for only one case, one could consider that the “most frequently traveled path” wins—chus the ‘outcome would be considered as having a [1] value for all nine cases Note, however, that this more probabilitc strategy is disputable from a “ease-oriented” perspective ‘Of course, the stravegy(es) chosen must be lustiied on empirical grounds (cate-based leiowledge) and/or on theoretical grounds and not be the resut of some opportunistic “manipulation.”50 CONFIGURATIONAL COMPARATIVE METHODS If none of these strategies, or a combination thereof, resolves the contra- dictory configurations, some cases will have to be removed from the key mi mization procedure. In such an event, there are basically four options: * Choose to move on and proceed with esQCA, even though there are still one or ‘more contradictory configuration() left. There are then two sub-options: Either delete the cases involved in the contradictions from the data table or keep them inthe data table." Those cases sill involved in the contradictory configuration(s) could then be interpreted separately (apart from the esQCA procedures proper), using a more qualitative-historical, case-specific approsch. ‘© Consider using mvQCA or fsQCA (see Chapters 4 and 5), which are able to process more fine-grained dita. Indeed, the reason esQCA easily produces con- tradictions is simply because dichotomization strongly reduces the richness of the data and hence also mases potential diferences across the cases (se p. 148: “costs” and “benefits” of dichotomization) ‘© Consider turing to other techniques, quantitative or qualitative. We recommend that {You atleast try out mvQCA and/or f6QCA first (or in parallel, so you can weigh the ‘rengths and limitations ofthe QCA techniques vis-bvis the other techniques), because these two other QCA techniques will alow you to keep some key sirengths of the QCA approach (both analytic and case-riented, ete. —see Chapter 1). ‘+ com if you are using esQCA for theory-testing, stop there and happily con- ‘lude that esQCA has allowed you to falsify the theory (see pp. 3,16) ‘Technically speaking, if the decision is to proceed with esQCA, itis neces- sary t0 produce a revised dichotomized data table, which enables the software to produce a revised truth table. In real-life research, experience shows that several iterations may be necessary to obtain a contradiction-free truth table. For this textbook example, we opt for the pragmatic way: Add a fifth con- dition to the model. In substance, we choose to add a “political-institutional” condition to the four more so:ioeconomic conditions derived from Lipset's ‘theory. This fifth condition is governmental stability (GOVSTAB). The thresh- old is placed as follows: A score of [0] (low stability) if 10 cabinets or more hhave governed during the period under investigation and a score of {1 (high stability) if fewer than 10 cabinets have governed during that same period. The addition of this fifth condition can be justified on theoretical grounds: For the cases of “breakdown of democracy": In the context of already less favorable socioeconomic circumstances, governmental instability further weakens the political system, the institutional capacity to address problems, tnd the credit of democratic institutions. Conversely, for the “survivors,” more stable governments are able to consolidate democratic institutions and enhance their capacity to confiont political challenges. CCRISP SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) st We thus obtain a new raw data table (Table 3.5, with one additional column 4s compared to Table 3.2), which is dichotomized (Table 3.6). The software then produces a new truth table (Table 3.7). Table 35 Lipset's casero | cnecar | ursawiza | crreracy | moras | covstaa | suavivat aus Ee o8 ot | 0 ° EL 098 | os vas | 489 4 1 cae us | os9 | ata 6 1 EST wes [es %5 uf «6 | o FN so | 2 vt | 9 1 FRA om | ma 962 |e 5 1 ER ms | 65 8 wos [ou a RE so [aa sa | mt | w ° mules | rea | es S 167] 6 ° RE oa | os %5 us 5 1 TTA si” [ana mi | 96 9 ° et | 1008 | 788 os | wa 2 1 POL 30 | om com ae o | POR 320 153, 38 ai 9 0 rom | al 219 ous | 7 ° SPA ar | se | as | 2 ° SWE om | om wo | ma | 6 i UK vos [7 9 | 99 ‘ 1 [Labels for conditions: same as Table 3.2, pls a ih condition: GOVSTAB: Governmental stability (umber of cabinets in pci) (Case abbreviations and sources same as Table 32.)2 CONFIGURATIONAL COMPARATIVE METHODS ‘Table 3.6 Lipset’s Indicators, Dichotomized Data, Plus a Fifth Condition (CRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) 3 ‘Table 3.7. Truth Table of the Boolean Configurations (4 + 1 Conditions) cass | anrcer | oman | urercr | monan | acuta | sav | ansci ane | e||see| oem [enn mw [ete fs tte]. 7 (sl seu ea |e |e | @ [oe topo pep pe mmo poe fp pe ps fe og |eecien pprong| ean pa a Se || * eo cacia| exon mrae| ece|zoms| eps |B ca [oe foe poe pe |e fe = ce exo gum eam more | ron tu am | mm ponte | an DC ITA o o 0 o 1 _ POL Seema ee oa ann |i] || = |e mee POL o o 1 o ROM rx [oe toe poe pe pe fe wow [oe tle pe pets | ‘SPA o o o o o 9 ‘Table 3.7), which have a [1] value on (GOVSTAB]. Some other contradictions: tvs Dor etled ne sue way : Tower, te sil icone conadicoy congtin, embracing (vo UK 1 1 1 1 1 1 cases: Estonia and Finland. Even with the addition of a fifth condition, those to caer alse the some wes onal cone nd yo they ily different outcomes: Estonia is a “breakdown” case ({0] outcome), whereas ‘This truth table (Table 3.7) icher” than the previous one (Table 3.4): By Finland is a “survivor” case ({1] outcome). In such a situation, we now envis~ ‘adding 4 condition, we move from 6 to 10 configurations, so indeed we have added diversity across the cases. This has enabled us to resolve most contra- dictions. Consider, for instance, the three cases of Austria, Sweden, and France, which formed a contradictory configuration when we considered only the four Lipset conditions (see first row in Table 3.4). By adding the [GOV- STAB] condition, we can now differentiate Austria (first row in Table 3.7), ‘which has a (0] value on [GOVSTAB], from Sweden and France (fifth row in age three possible options, The first one would be to further reexamine the ‘model, which could—possibly—lead to the inclusion of a sixth condition (good practice” strategy 1 in Box 3.6), The problem, however, is that the model becomes more complex with the addition of each condition and less clear for the pedagogical purpose of this textbook. The second option is simply ‘to accept that those two cases deserve some specific qualitative-historical inter- pretation and that hence they should be left out for the next steps of the esSQCA.s CONFIGURATIONAL COMPARATIVE METHODS ‘The third option (“good practice” strategy 3 in Box 3.6) isto reexamine the way in which the various conditions included in the model have been opera tionalized, with a particular focus on the cases of Finland and Estonia. Doing this, we discover that if we move the threshold of the GNPCAP condition from, ‘$600 to $550 (actually this latter threshold is located near a natural “gap” in the data!'), this allows us to differentiate between Finland ($590) and Estonia (S468), Incidentally, note that this modification of the threshold also changes the score for Czechoslovakia ($586: from a [0] 10 a [1] value). More impor- tant, it allows us to produce a contradiction-free truth table,” as is shown in the next two tables (Tables 3.8 and 3.9) ‘Table 38 _Lipset’s Indicators, Dichotomized Data, Plus a Fifth Condition (and GNPCAP Recoded) T T caseip | cwecar | uesanza | umeracy | iwo.as | Govstas | suRvVAL aus i ° 1 1 ° ° BEL, T 1 7 7 T T CE o 1.0 T T T EST 0 0 1 IN 7 0 1 FRA 7 a T T i 7 GER 7 r t T GRE 0 HUN o TRE 1 TTA, 0 POL POR } | t j ‘CRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) 5s ‘Table 39 ‘Truth Table of the Boolean Configurations (4+ 1 Conditions, GNPCAP Recoded) casew | avecar | ursaniza | urrekacy | iwotas | covstae | SURVIVAL aus q 0 1 1 0 0 BEL, t 1 1 1 1 r cze, NET, UK EST 0 0 1 0 1 ° FRA, 1 ° 1 1 1 qi SWE GER 1 1 ' 1 0 ° GRE, ° ° 0 ° ° ° Por, SPA o ° t ° ° o L ° 1 ° 1 1 mA, o ° ° o ' o ROM Table 3.9 can also be grasped more visually, through a Venn diagram (Figure 3.3). This five-dimensional diagram is a bittess easy to grasp than the previous, four-dimensional Venn diagram (Figure 3.2, above), but itis built on the same premises: Each condition still spits the logical space into two equal parts (of 16 basic zones each). Graphically, what is new is that the visualiza- tion of the fifth condition (GOVSTAB) requires two separate “patches” (two horizontal squares, each one comprising 8 basic zones, in which this condition hhas a [1] value), Note also that many more basic zones of the logical property space are left empty (as compared with the previous, four-dimensional dia- gram; Figure 3.2)—a reminder that the more conditions we include ia the ‘model, the more limited the observed empirical diversity (see p. 27), Indeed the revised, contradiction-free truth table (Table 3.9) now places Finland and Estonia in two separate configurations. More precisely: Estonia is now alone in a specific configuration, and Finland has joined Ireland in another configuration. Note also that Czechoslovakia has also moved and joined the56 CONFIGURATIONAL COMPARATIVE METHODS eo] “GRE, POR, SPA amo] — an nots] : ao0aT om aor fa "TA OM, scbtene SOI) [oa fooro] [00770] ra , ferioo]_ [oro] ER Trivia] [11100 , (ore0r] oon [ror [oor g om fr [ox000] [77000} ee igure 33 Venn Diagram (5 Conditions: += Venn diagram produced by the “visualizes” tol, TOSMANA 1.3.0.0 software. configuration of “perfect” survivor cases ([1] values on all conditions, leading, toa [I] value on the outcome): Belgium, the Netherlands, and the UK. STEP 4: BOOLEAN MINIMIZATION For this key operation of esQCA, the material used by the software is the truth table (Table 3.9) with its nine configurations: three configurations with a (1 out come (corresponding to 8 cases), and six configurations with a (0} outcome (cor- responding to 10 cases). As.is obvious by looking at Table 3.9, each configuration 'may correspond to one or more empirical cases (or to none—the “logical remain- der” configurations; see p. 59). What is important to mention here is, that the software does not recognize cases but rather the configurations specified in the truth table, Thus the number of cases in each configuration will not be relevant in the course of the minimization process. After the minimization, however, it will be possible to connect each of the cases to the minimal formula that is obtained. CCRISP.SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) 7 ‘The software minimizes these configurations, using the Boolean minimiza- tion algorithms (see Box 3.2), considering separately the [1] configurations and the [0] configurations. One must thus apply the minimization procedure ‘vice, first for the [1] configurations, and then for the [0] configurations. The sequence is not important, as Tong as both are carried out. It is important to minimize both types of configurations, because we do not expect to find some form of perfect “causal symmetry” in social phenomena (see p. 9). In other words, we should not deduce the minimal formula for the 0] outcome from that of the [1] outcome, or vice versa, although it would be technically feasi- ble in some circumstances to do so by applying the De Morgan's law." ‘Minimization of the [1] Configurations (Without Logical Remainders) First, we ask the software to minimize the (1] configurations, without including some non-observed cases. We obtain the following minimal formula (Formula 1): GONPCAP*LITERACY* + GNPCAP utunza* 9 SURVIVAL INDLAB*GOVSTAB "LITERACY = GOVSTAB (BEL, CZE,NET.UK + (FIN IRE + FRA, SWE) FRASWE) | ‘This is called a “descriptive” formula, because it does not go much beyond the observed, empirical cases. In consists of two terms, each one of which is a combination of conditions linked with the « 1 » outcome value. Following the Boolean notation (see Box 3.1), it can be read as follows: “The ‘1” outcome (survival of democracy) is observed: ‘© In countries that combine high GNP per eapita (GNPCAP] AND high literacy rates [LITERACY] AND high percentage of industrial labor forces (INDLAB] AND high governmental stability [GOVSTAB] oR + In counties that combine high GDP per capita [GNPCAP] AND low urbaniza- tion {urbaniza] AND high literacy rates [LITERACY] AND high govermmental stability (GOVSTAB]" The first term of the minimal formula corresponds to six countries: on the one hand Belgium, Czechoslovakia, the Netherlands, and the UK (which share the same configuration—ice., with all five conditions) and on the other hand!8 CONFIGURATIONAL COMPARATIVE METHODS France and Sweden, The second term corresponds to four countries: on the one hand Finland and Ireland and on the other hand France and Sweden. Note that concerning France and Sweden, we thus have two partly “concurrent explanations, In such a situation—which is quite often met with esQCA—the researcher has to make a choice, using his or her case knowiedge. This is part Of the phase of interpretation of the minimal formula (see Step 6, p. 65). ‘This descriptive minimal formula is still quite complex, as each term still, includes four out of the five conditions of the model. Only a small measure of parsimony has been achieved. The formula does allow some first interpreta- tions, however. For instance, we could interpret the fact that the (URBANIZA] condition does not play a role in the survival of democracy in Belgium, Czechoslovakia, the Netherlands, and the UK. [Note that the two terms of the formula share the [GNPCAP * LITERACY * GOVSTAB] combination of conditions. Thus, this combination can be made more visible by manually modifying the minimal formula (this is not done by the software). We simply treat the formula as a conventional algebraic expres- sion, a sum of products, and factor out the common conditions. This will pro- duce a more structured versien of the formula—not a more parsimonious one, because no further conditions are eliminated by this operation. Thus, Formula, 1 can be rewritten as follows (Formula 2) INDLAB envearstiteracy scovstan + { . SURVIVAL, ‘This rewriting of the formula shows quite clearly what is common to all the “survivor” eases (the left-hard side of Formula 2)—once again, this could be subject to interpretation by the researcher, as indeed this core combination of three conditions is shared by all “survivor” cases. The rewritten formula also shows what is specific to each one of the two clusters of cases (the two differ- cent “paths” on the right-hand side of Formula 2). “Minimization of the [0] Configurations (Without Logical Remainders) Secondly, we perform exactly the same procedure, this time for the [0] con. figurations and also without including some non-observed cases. We obtain the following minimal formula (Formula 3) (CRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) 2 aanpeap * urbaniza* indlab + (EST + GRE, POR, SPA HUN, POL + ITA, ROM) GNPCAP * LITERACY * > INDLAB * govstab survival (AUS + GER) A the previous one, this minimal formula is also quite complex. Reading this formula (following the same conventions as explained above), we see that esQCA provides us with two “paths” to the [0] outcome. The first one corre- sponds to many cases: Estonia; Greece, Portugal, and Spain; Hungary and Poland; Italy and Romania. These eight cases of democracy breakdown all share the [gnpcap * urbaniza * indlab] combination—ice., the combination [0] values on three conditions—which is quite consistent with the theory. The second one is specific to Austria and Germany—note that tis result could have been guessed by looking at the Venn diagram (Figure 3.3, p.56), in which those two cases are “distant” from the eight above-mentioned cases. This formula ceannot be rewritten in “shorthand” manner, because the two terms have noth- ing in common, STEP 5: BRINGING IN THE “LOGICAL REMAINDERS” CASES. Why Logical Remainders Are Useful The problem with Formulas 1 to 3 is that they are still quite complex: Relatively litle parsimony has been achieved. To achieve more parsimony, it is necessary to allow the software to include non-observed cases, called “log- ical remainders.” In this example, remember that there isa large “reservoie” of logical remainders as is seen in the Venn diagram (Figure 3.3, p. $6). Only a tiny proportion of the logical property space is occupied by empirical eases: (Out of the 32 potential configurations (=2,as there are 5 conditions: see p. 27), only 9 correspond to observed cases. Thus, the 23 logical remainders (= 32 minus 9) constitute a pool of potential cases that can be used by the software to prosiuce « more parsimonious minimal formula ‘Why does the inclusion of logical remainders produce more parsimonious minimal formulas? This can be explained visually, using the Venn diagram and eight concrete cases: all those eases with a [0] outcome, which also happen to be situated on the left-hand side of the Venn diagram: (Estonia; Greece, Portugal and Spain; Hungary and Poland; Italy and Romaaia.o CONFIGURATIONAL COMPARATIVE METHODS First, note that the simpler (the “shorter”) a Boolean expression, the larger the number of configurations it covers: * A combination of all five conditions covers only one configuration (eg, the [00000} zone, which contains the Greece, Portugal, and Spain eases), + A combination of four conditions covers two configurations: If we want 10 “cover” not only the Greece, Portugal, and Spain cases but also the zone right near it that contains the Taly and Romania cases (the {00001} zone), we only need to have information about four conditions. Indeed, we don’t need t0 know about the [GOVSTAB] condition: It has a [1] value for the Italy and Romania ‘eases and a [0] value for the Greece, Portugal, and Spain cases + Likewise, a combination of three conditions covers four configurations. # Accom! nation of two conditions covers eight configuration, ‘+ And a statement containing only 1 condition covers 16 configurations —ie., half of the Boolean propery space, For instance, the 2one corresponding to a (0] value on the [GNPCAP] condition is the whole left half of the Venn diagram, corresponding to 16 configurations (only 4 of whieh, incidentally, contain some observed cases—those 8 cases that all happen to have a [0] outcome), Following this logic, the usefulness of logical remainders is quite straight- forward: To express those eight cases in a simpler way, it suffices to express them as part of a broader zone, also comprising some logical remainders. Hence what we can do is to make a “simplifying assumption” regarding the 12 logical remainders on the left-hand side of the Venn diagram: Let us assume that, if they existed, they would also have a [0] outcome, jus like the 8 observed cases. If this assumption is correct, then we have produced a much larger zone (the whole lefi-hand side of the Venn diagram, comprising 16 configurations) sharing the (0) ‘outcome, and thus the 8 observed cases can be expressed in @ much more parsi- ‘monious way: simply [gnpcap|—i., [0] value for the [GNPCAP] condition. This is exactly what the software does: It selects some logical remainders (only those that are useful to obtain a shorter minimal formula), adds them to the set of observed cases, and makes “simplifying assumptions” regarding these logical remainders. This then produces a simpler term in the minimal formula. ‘Minimization of the [1] Outcome (With Logical Remainders) Running again the minimization procedure, this time allowing the software to include some of the logical remainders, we obtain the following minimal formula (Formula 4): GNPCAP * GOVSTAB > SURVIVAL (BEL, CZE, NET, UK + FIN, IRE + FRA, SWE) CCRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) o It is read as follows: “For all these countries, a high GDP per capita, com- bined with governmental stability, has led to the survival of democracies in the inter-war period.” Comparing this formula with Formula 1 (see p. 57), we see that a more parsimonious solution has been achieved, thanks to the simplify- ing assumptions made by the software regarding some of the logical remain- ders. We can obtain a list of these simplifying assumptions from the software and lay them out in the report of the analysis—in this instance five of them were used’ V/ GNPCAP( 1) URBANIZA (0)LITERACY (0)INDLAB (0}GOVSTAB(1] 2/ GNPCAP(1) URBANIZA (0)LITERACY {O}INDLAB {1}GOVSTAB{1] 3/ GNPCAP(1] URBANIZA 1] LITERACY {0}INDLAB{0}GOVSTAB( 1) 4) GNPCAP (1) URBANIZA 1] LITERACY{0}INDLAB {1 }GOVSTAB(1) 5/ GNPCAP(1] URBANIZA1)LITERACY {1 INDLAB {0}GOVSTAB(1} ‘These simplifying assumptions can be visualized in the Venn diagram (through the TOSMANA software). In Figure 3.4, the minimal formula (the ‘solution”) is represented by the horizontal stripes. This area corresponds to the three configurations with observed cases displaying a [1] outcome, plus five logical remainder configurations, ‘Minimization of the [0] Outcome (With Logical Remainders) Likewise, we obtain the following minimal formula (Formula 5): ‘anpeap + govstab > survival (EST+GRE,POR, SPA (AUS + GER + GRE, POR, HUN, POL + ITA, ROM) SPA + HUN, POL) It is read as follows: ‘© “In cight countries (Estonia, ...and Romania), low GNP per capita ‘explains’ the breakdown of democracy in the inter-war period # In seven countries (Austria, .. and Poland, governmental instbil the breakdown of democracy inthe inter-war period.” y ‘explains ‘There are thus two alternative paths toward the [0] outcome. Note that for five countries (Greece, Portugal, and Spain; Hungary and Poland), both pathse CONFIGURATIONAL COMPARATIVE METHODS "GRE, POR, SPA. foooro ft T0030 aINDUAS foooor Coon Too PITA, ROM corer] [oor a sa0vsTAB: feotoo] oor Horo HON, POL| scovsms [ovios) [ori oon Gon raw Figure 34 Venn Diagram: Solution for the {1] Outcome (With Logical Remainders)* + Venn diagram produced by the “suze” tool, TOSMANA 1.3.0.0 software are valid. In such a situation, the researcher must choose, country by country, and relying on his or her case knowledge, which path makes more sense. ‘Comparing this formula with Formula 2 (see p. 58), we see that substantial parsimony—even more" than in Formula 4—has been gained, thanks to the “simplifying assumptions” made, by the software, regarding some of the log- ical remainders. We can also obtain a lst of these simplifying assumptions and lay them out in the report of the analysis. In this ease, many more have been used (18 in all). This can be visualized through a Venn diagram (Figure 3.5; same conventions as Figure 3.4, Examining Figures 3.4 and 2, itis clear that the minimal formula for the {0} outcome (including 18 logical remainders) and the minimal formula for the [1] outcome (including five logical remainders) are the perfect logical complement of one another. In other words, the software has used up all the available “empty space” of logical remainders, so as to produce the most CCRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) 6 Sora OT0 = ra “enpcaP Minimizing: 0 including: R Figure 35 Venn Diagram: Solution forthe 0] Outcome (With Logical Remainders)* + Venn diagram produced by the “isuliee” ool, TOSMANA 1.30.0 software parsimonious minimal formulas possible. Thus, the wo minimal formulas completely fill up the Boolean property space, well beyond the observed cases. (This use of logical remainders raises a concern and also an important tech- nical issue. First the concern: Isn’t it altogether audacious to make assump- tions about non-observed cases? One way to frame this debate is to question the relative plausibility of those simplifying assumptions. This issue is addressed in detail in Ragin and Sonnett (2004; see also Ragin, 2008), in our review of the critiques of QCA (see p. 152) and of applications (sce p. 135), and is exemplified in the fsQCA application below (cee pp. 110-118), For reasons of space, we cannot engage here in a full discussion of this issue, using this inter-war project data. The key point to remember is that itis always possible to restricr the choice of logical remainders used by the software, If we do this, we obtain somewhat less parsimonious minimal formulas.“ CONFIGURATIONAL COMPARATIVE METHODS Second, the issue: What if the software uses the same logical remainders both for the minimization of the [1] configurations and for the minimization of the [0] configurations? If it does, it will produce “contradictory simplifying assumptions,” because indeed it would be a logical contradiction to assume that a given (non-observed) case would simultaneously have a {1} outcome and a {0] ‘outcome, Gladly, inthis example, its not the case. The comparison of Figures 3.4 ‘and 3.5 shows that the two minimal formulas (with logical remainders) do not overlap. If we had encountered such a difficulty, it would have been possible to address it, sing more advanced technical steps (see p. 136). Finally, note that itis useful at this stage to assess the coverage of the min- imal formulas—that is, the way the respective terms (or “paths") of the ‘minimal formulas “cover” the observed cases. This is a second measure of the “fit” of the model, as the measure of consistency (see p. 47) at a previous stage, Technically, one should make three measures, both for the [1] and [0] ‘outcome values. For instance, for the [1] outcome value: (a) raw coverage: the proportion of [1] outcome cases that are covered by a given term; (b) unique coverage: the proportion of [1] outcome cases that are uniquely covered by a given term (no other terms cover those eases); (€) solution coverage: the pro- portion of cases that are covered by all the terms.” Poe “Good Practices" (6): Four Complete Minimization Procedures to Be Run and Made Explicit CChecklst for the minimization procedure(s), using the computer sofware: ‘= Perform the minimization. both with and. without Ineusion of logical remainders. Each of these approaches may yield information of some interest. “Thus: Four complete minimization procedures must be run: © [1] configurations, without logical remainders © [1] configurations, with logieal remainders © [0 configurations, without logical remainders, © [0] configurations, wih logical remainders ‘Ack the software to list the “simplifying assumptions” and display those in your research report. i Check for possible “cantradictory simplifying assumptions” and, Insofar a6 possible, solve ther (see p. 136). CCRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) 6 ‘Present al your minimal formulas (including the case labels). and if needed use visual displays (eg, Venn diagrams) to make the minimal formulas more understandable forthe reader. ‘If is useful for your Interpretation factor (by hand) some conditions, to make the key regularities in the minimal formulas more apparent. ‘+ Assess the “coverage” ofthe minimal formulas—i.e.,the connection between the respective terms of the minimal formulas and the observed cases. STEP 6: INTERPRETATION Remember that as a formal data analysis technique, but even more so because itis a “case-oriented” technique (see p. 6), «sQCA (the formal, eomputer-run par of it), as wel asthe other QCA techniques, is not an end in itself; rather, itis a tool to enhance our comparative knowledge about cases in small- and imermediate-N research designs. ‘This means that the final step of the procedure is a crucial one: The researcher interprets the minimal formulas. The emphasis can be laid more on theory or on the eases, or on both, depending on the research goals. Obviously this requires a “return to the cases” using the minimal formula(s) that is (are) considered most relevant. In the inter-war project example we have unfolded inthis chapter, some case-based interpretations could follow from questions such as the following: What is the “narrative” behind the fact that, according to the minimal formula, high GNP per capita combined with governmental stability has lead to survival (or non-breakdown) of democracy in counties such as Belgium, Czechoslovakia, the Netherlands, and the UK? Is the “causal” story the same in these four countries? What distinguishes them from other countries also covered by this same minimal formula, such as France and Sweden? Why does low GNP per capita, asa single factor, seem to play a more prominent role in the breakdown of democracy in countries such as Estonia, Italy, and Romania? Conversely, why does a more directly “political” factor (eovernmental instability) come out as the single key determinant in democ- racy breakdown in countries such as Greece, Portugal, and Spain, even though these are also relatively poor countries (low GNP per capita as well)? To what extent is the “narrative” behind the German and Austrian cases of democracy breakdown really comparable? And so on ‘To sum up: esQCA minimal formulas allow the researcher to ask more focused “causal” questions about ingredients and mechanisms. producing66 CONFIGURATIONAL COMPARATIVE METHODS (or not) an outcome of interest, with an eye on both within-case narratives and cross-case patterns, Note that unless individual conditions can be clearly sin- agled out (e.g, the condition is clearly a necessary condition or comes close to being a necessary and sufficient condition), it is important to refrain from interpreting relations between singular conditions and the outcome. At the interpretation stage, do not lose sight of the fact that the richness of esQCA. minimal formulas resides precisely in the combinations and “intersections” of conditions. It would be a pity to lose a chance to gain some “configurational knowledge” (Nomiya, 2004) at this crucial stage. Note, finally, that these rules, | and “good practices’ of interpretation also apply to mvQCA and fsQCA. ‘= csQCA is based on a spocic language, Boolean algebra, which uses only binary data ((0] or [1}) and is based on a few simple logical operations. ‘As any language is convencions must be properly used les a formal, but non statistial, language. ‘= Ieis importane ro follow a sequence of steps, from the construction of a binary data table to the ial "minimal formula '» Two key challenges in this Sequence, before running the minimization procedure, are: (1) implementing a useful and meaningful dichotomi- zation of each variable and (2) obtaining a “truth table” (table of configurations) that is ee of “contradictory configurations.” ‘+The key esQCA procedure is"Boolean minimization” One must run the minimization procedures both for the [1] and the [0] outcomes, and both with and without ‘logical remainders” (non-observed cases) ‘The use of logical remainders"—and the “simplifying assumptions” that are made on them by the sofeware—rases some principal and technical ificutes, bu che later can be addressed. ‘= Obtaining minimal formulas is only the end of the computeraided part fof esQCA It marks the beginning ofa key final step-case- andlor theory- informed interpretation, which should be focused on the link between key ‘combinations of conditions and the outcome. ‘ Hai Key Complementary Readings CCaramani (2008), De Meur & Rihoux (2002). Ragin (1987, 2000), ‘Schnelder & Wagemann (2007, forthcoming). CCRISP-SET QUALITATIVE COMPARATIVE ANALYSIS (CSQCA) o NOTES Note that, in most publications so far the equal symbol [=] has been used. Following Schneider and Wagemann's (2007) suggestion, we recommend using the nbol [9] instead. One of the key reasons therefor is that, by using thi 5ym- bol the Boolean formula Gan not be mistaken witha standard statistical (eg. epres- sion) equation, 2 cually, inthis example, we can't tl, because we should also examine the (combinations 60) conditions lesdng tothe absence ((0] vale) ofthe outcome. 3. A condition comesponds tan “independent variable” in staisical analysis, However, itis not an “independent” variable in the statistical sense. There is no assumption of independence between the conditions—quite the contrary; we would txpect combinations tobe relevant (se Chapter 1). “For more detailed justifications, se Berg-Schloser and De Mear, 1994. 5. Typically, one should not use the median or the mean if this locates the dichotomtzation threshold in an area ofthe data distribution where many cases are Situated (his would give opposite scores [0 or 1] to cass that display quite proximate ‘aw values) 6 Actually the software (¢, TOSMANA) can already be used in the prior stages—for example, forthe clustering of cass (dchotomization) if one uses more technical ereria “7 Itis also a problem for further QCA software treatment. Each time the software meets a“don't care” configuration it wil produce two distinct configurations: one with 2 [0] outcome and one wth a {1} outcome, This might not make sense from an emp ical perspective. In practice the “don't cae” outcome i aly used, and even then i is used simply to signal a combination of conditions that is empirically impossible (eg. pregnant males). Tis actualy more complex—depending on whether or not the cates involved in those contradictory configuration are left (by the researcher) inthe data able this will fave an impact on the sizeof he “eeervir” of non-observed “Topical remainder” cases that canbe used bythe software in the minimization procedue (see note 10, below) 9. See also futher discussion on p. 132 10, Note that if one chooses to include logical remainders, these two sub-options will have a diferent influence onthe end result (the minimal formals), Ifthe eases involved in contradictions are deleted ought, the logieal space they used 10 occupy til be left “open: and the software wil have the possibility fo use ths “fee space” in its seach for useful logical remainders—ths is Hiely to generate a shorter, more pl simonious minimal formula, Conversely if one keeps those cases in the data table, the Tosial space will be “occupied” by those cases, andthe reservoir of potential “logical femainders” wl be a itl more constsined~—tissUely to generale a less pasimo- tous minimal forma, Probably itis advisable to prefer the second, more cautious option, because we cannot assume thatthe logical space occupied by these conradi- tty cassis “expt expiical epoca”“ ‘CONFIGURATIONAL COMPARATIVE METHODS 31, Thiaca be vied ning the "rele" fecon in TOSMANA ne P.79), 12, This key suggestion has been made by Svend-Erik Skaaning, 13, For mary reasons, one of which being that social phenomena display Tinted diversity (for a more detailed discussion, see De Meur & Rihoux, 2002). See’ so pp. 8-10 on the issue of causal complexity. However, under some very spetific circumstances, De Morgan’s law can be meaningfully applied (see Wagemann & Schneider, 2007, p. 26). 14, In the TOSMANA output, the “+” sign in between (groups of eases separates ceases with different configurations (in the truth table) 15, ‘The notation used here is that of the TOSMANA software for simplifying ‘assumptions. GNPCAP(1} simply means [1] value on the GNPCAP condition; and so ‘on (ee Box 4.1). 16. This is simply because even more logical remainders have been included by the software for the minimization procedure, 17. For mote details, see Ragin (2006b, 2008), Goertz (2006b), and Schneider & Wagemann (2007, fortheoming). Note that this operation is also implemented in myQCA and FSQCA. Multi-Value QCA (mvQCA) Lasse Crongvist Dirk Berg-Schlosser ‘After reading thie chapter you should be able to do the following: ‘= Reflect on the limitations of using dichotomized coneltions and on the potential advantages of using multi-value conditions ‘+ Read and use the specfic mvQCA notation: grasp the basis of muli- value minimization ‘+ Make informed choices regarding threshold values 's Replicate a standard mvQCA procedure, step by step, using the software (TOSMANA) ‘= Reflect on the respective strengths and limitations of both esQCA and ‘mvQCA for your own research WHY myQca? Some Problems in the esQCA Example Remember that in the previous chapter on esQCA we quickly bumped into a first difficulty: the presence of many contradictory configurations. To solve these, as a technical solution, we had to add a fifth condition related to gov- cemmental stability. However this technical solution still leaves some problems unresolved, or at least is not fully satisfactory, in at least three respects. First, by adding this fifth condition, we have moved beyond the simple testing of Lipset’s theory, because we have included in the model a condition that is outside the realm of that theory. Second, to obtain sufficiently parsimonious minimal formulas, we «70 ‘CONFIGURATIONAL COMPARATIVE METHODS included a very lange number of “logical remainders,” some of which could be questioned in terms of plausibility. Third, when we chose nor to consider the logical remainders, we obtained more “descriptive” minimal formulas that, with csQCA at least, were not sufficiently parsimonious. ‘Actually, all these limitations may stem in part from the fact that the con-

You might also like

• Methods in Analytical Political Theory Compress

  

  No ratings yet

  Methods in Analytical Political Theory Compress

  354 pages
• Suggestions For A Sociological Approach To The Theory of Organizations

  

  No ratings yet

  Suggestions For A Sociological Approach To The Theory of Organizations

  16 pages
• Implementación

  

  No ratings yet

  Implementación

  188 pages
• Seeking An Understanding of Poverty That Recognizes Rural-Urban Differences and Rural-Urban Linkages

  

  No ratings yet

  Seeking An Understanding of Poverty That Recognizes Rural-Urban Differences and Rural-Urban Linkages

  19 pages
• Meguid 2008 Party Competition Between Unequals

  

  No ratings yet

  Meguid 2008 Party Competition Between Unequals

  338 pages
• Carl Rogers

  

  No ratings yet

  Carl Rogers

  18 pages
• The New Transnationalism: Klaus Dingwerth

  

  No ratings yet

  The New Transnationalism: Klaus Dingwerth

  275 pages
• Social Science Data Analysis an Introduction 3658412291 9783658412296 Compress

  

  No ratings yet

  Social Science Data Analysis an Introduction 3658412291 9783658412296 Compress

  194 pages
• Between X and Y How Process Tracing Contributes To Opening The Black Box of Causality

  

  100% (1)

  Between X and Y How Process Tracing Contributes To Opening The Black Box of Causality

  19 pages
• Lectura 3 Bachrach - and - Baratz - 1962 LAS DOS CARAS DEL PODER

  

  No ratings yet

  Lectura 3 Bachrach - and - Baratz - 1962 LAS DOS CARAS DEL PODER

  7 pages
• IMBENS e RUBIN. Rubin causal model

  

  No ratings yet

  IMBENS e RUBIN. Rubin causal model

  15 pages
• Best, Cotta - 2000 - Parliamentary Representatives (Desprotegido)

  

  No ratings yet

  Best, Cotta - 2000 - Parliamentary Representatives (Desprotegido)

  54 pages
• WILDAVSKY Aaron. Speaking Truth To Power PP 41-61

  

  No ratings yet

  WILDAVSKY Aaron. Speaking Truth To Power PP 41-61

  11 pages
• Richard S. Katz Bernhard Wessels The European PBook4You

  

  No ratings yet

  Richard S. Katz Bernhard Wessels The European PBook4You

  295 pages
• Amartya Sen Approach in Human Development

  

  No ratings yet

  Amartya Sen Approach in Human Development

  17 pages
• Diarmuid Maguire Marxism 2010

  

  100% (1)

  Diarmuid Maguire Marxism 2010

  10 pages
• Martha K. Huggins, Mika Haritos-Fatouros, Philip G. Zimbardo - Violence Workers - Police Torturers and Murderers Reconstruct Brazilian Atrocities (2002) PDF

  

  No ratings yet

  Martha K. Huggins, Mika Haritos-Fatouros, Philip G. Zimbardo - Violence Workers - Police Torturers and Murderers Reconstruct Brazilian Atrocities (2002) PDF

  317 pages
• Qualitative Methods and The Study of Civil War Course Syllabus

  

  100% (1)

  Qualitative Methods and The Study of Civil War Course Syllabus

  6 pages
• (Luc Boltanski, Laurent Thevenot, Catherine Porter

  

  No ratings yet

  (Luc Boltanski, Laurent Thevenot, Catherine Porter

  203 pages
• BUZAN, Barry From International To World Society 2004

  

  No ratings yet

  BUZAN, Barry From International To World Society 2004

  320 pages
• Comparative Historial Analysis in The Social Sciences. How Institutions Evolve Insights From Comparative Historical Analysis, 208 - 240

  

  100% (2)

  Comparative Historial Analysis in The Social Sciences. How Institutions Evolve Insights From Comparative Historical Analysis, 208 - 240

  451 pages
• Bruce Ackerman PDF

  

  No ratings yet

  Bruce Ackerman PDF

  10 pages
• The Spaces of Democracy and The Democracy of Space A New Network Exploring The Disciplinary Effects of The Spatial Turn

  

  100% (1)

  The Spaces of Democracy and The Democracy of Space A New Network Exploring The Disciplinary Effects of The Spatial Turn

  7 pages
• Vivien Schmidt 2012 Discursive Institutionalism

  

  No ratings yet

  Vivien Schmidt 2012 Discursive Institutionalism

  25 pages
• Arthur L. Stinchcombe - The Logic of Social Research (2005, University of Chicago Press)

  

  No ratings yet

  Arthur L. Stinchcombe - The Logic of Social Research (2005, University of Chicago Press)

  184 pages
• Luc Boltanski - Love and Justice As Competences-Polity (2012)

  

  No ratings yet

  Luc Boltanski - Love and Justice As Competences-Polity (2012)

  351 pages
• Skvoretz & Fararo (2011) - Mathematical Sociology

  

  No ratings yet

  Skvoretz & Fararo (2011) - Mathematical Sociology

  14 pages
• Briefing Paper: Electoral Corruption

  

  No ratings yet

  Briefing Paper: Electoral Corruption

  12 pages
• Criminal Governance Mantilla Feldmann (1) (1)

  

  No ratings yet

  Criminal Governance Mantilla Feldmann (1) (1)

  25 pages
• Social Origins of Institutional Weakness and Change Preferences Power and Police Reform in Latin America

  

  100% (1)

  Social Origins of Institutional Weakness and Change Preferences Power and Police Reform in Latin America

  44 pages
• Chapter 2 PDF

  

  No ratings yet

  Chapter 2 PDF

  35 pages
• Schneider 2019

  

  100% (1)

  Schneider 2019

  19 pages
• Comparative Electoral Management - Performance - Networks and Instruments Toby James 2019 Routledge 97

  

  100% (1)

  Comparative Electoral Management - Performance - Networks and Instruments Toby James 2019 Routledge 97

  416 pages
• (Social Indicators Research Series 70) Maggino, Filomena - Complexity in Society - From Indicators Construction To Their Synthesis (2017)

  

  No ratings yet

  (Social Indicators Research Series 70) Maggino, Filomena - Complexity in Society - From Indicators Construction To Their Synthesis (2017)

  334 pages
• Falleti and Cunial - 2018

  

  No ratings yet

  Falleti and Cunial - 2018

  88 pages
• 1 CRUZ, J. M. State and Criminal Violence in Latin America

  

  No ratings yet

  1 CRUZ, J. M. State and Criminal Violence in Latin America

  22 pages
• Criminal Violence in Latin America by Nicholas Barnes

  

  100% (1)

  Criminal Violence in Latin America by Nicholas Barnes

  10 pages
• What Are Microfoundations?

  

  No ratings yet

  What Are Microfoundations?

  18 pages
• John Mearsheimer - The False Promise of International Inst 1

  

  No ratings yet

  John Mearsheimer - The False Promise of International Inst 1

  46 pages
• Pandemic, Social Unrest, and Crime in U.S. Cities

  

  No ratings yet

  Pandemic, Social Unrest, and Crime in U.S. Cities

  19 pages
• Rossana Schanzer - Compressed

  

  No ratings yet

  Rossana Schanzer - Compressed

  17 pages
• Get Statistics A Tool For Social Researchers 4 Canadian Edition Riva Lieflander PDF Ebook With Full Chapters Now

  

  100% (7)

  Get Statistics A Tool For Social Researchers 4 Canadian Edition Riva Lieflander PDF Ebook With Full Chapters Now

  62 pages
• Schmidt 2010 PDF

  

  No ratings yet

  Schmidt 2010 PDF

  25 pages
• Sartori Comparing and Miscomparing Kopi

  

  100% (1)

  Sartori Comparing and Miscomparing Kopi

  15 pages
• Callen Institutional Corruption and Election Fraud Evidence From A Field Experiment in Afghanistan

  

  No ratings yet

  Callen Institutional Corruption and Election Fraud Evidence From A Field Experiment in Afghanistan

  41 pages
• Fischer2015 PDF

  

  100% (1)

  Fischer2015 PDF

  535 pages
• dp96 6 PDF

  

  No ratings yet

  dp96 6 PDF

  30 pages
• Surel - 2000 - The Role of Cognitive and Normative Frames in Policy-Making

  

  No ratings yet

  Surel - 2000 - The Role of Cognitive and Normative Frames in Policy-Making

  19 pages
• Schmidt, V.A. Discursive Institutionalism. The Explanatory Power of Ideas and Discourse

  

  No ratings yet

  Schmidt, V.A. Discursive Institutionalism. The Explanatory Power of Ideas and Discourse

  26 pages
• Electoral Violence - An Introduction

  

  100% (1)

  Electoral Violence - An Introduction

  12 pages
• Campbell, Converse, MIller & Stokes - The American Voter - 1960

  

  No ratings yet

  Campbell, Converse, MIller & Stokes - The American Voter - 1960

  12 pages
• New Perspective On Latin America

  

  No ratings yet

  New Perspective On Latin America

  9 pages
• State Formation - Political Science - Oxford Bibliographies

  

  No ratings yet

  State Formation - Political Science - Oxford Bibliographies

  28 pages
• The Historical Foundations of Symbolic Interactionism

  

  100% (1)

  The Historical Foundations of Symbolic Interactionism

  21 pages
• Esteban Ruizponce Madrid - Defining Institutions As Rules

  

  No ratings yet

  Esteban Ruizponce Madrid - Defining Institutions As Rules

  27 pages
• Configurational Comparative Methods

  

  No ratings yet

  Configurational Comparative Methods

  2 pages
• Patrick-A, Mello, Qualitative Comparative Analysis (Chapt. 2 I 8)

  

  No ratings yet

  Patrick-A, Mello, Qualitative Comparative Analysis (Chapt. 2 I 8)

  354 pages
• WagemannSchneider2007

  

  No ratings yet

  WagemannSchneider2007

  35 pages
• 1-s2.0-S2405844024074887-main

  

  No ratings yet

  1-s2.0-S2405844024074887-main

  40 pages
• Comparative Qualitative

  

  No ratings yet

  Comparative Qualitative

  2 pages
• Evaluating Organizational Change: How and Why?

  

  No ratings yet

  Evaluating Organizational Change: How and Why?

  29 pages
• Ajayi - Introduction To Research Methods - March 2023 - 065310 (1) 2

  

  No ratings yet

  Ajayi - Introduction To Research Methods - March 2023 - 065310 (1) 2

  61 pages
• IDENTIFYING A RESEARCH PROBLEM Summary'

  

  No ratings yet

  IDENTIFYING A RESEARCH PROBLEM Summary'

  14 pages
• Use Rubrics To Make Assessment Consistent Across Multiple Markers

  

  No ratings yet

  Use Rubrics To Make Assessment Consistent Across Multiple Markers

  2 pages
• Educational Leadership

  

  No ratings yet

  Educational Leadership

  5 pages
• SIR CLEO - Statistics-Exam

  

  No ratings yet

  SIR CLEO - Statistics-Exam

  11 pages
• Final Examination in Major 3214 - Language Research: Table of Specification

  

  No ratings yet

  Final Examination in Major 3214 - Language Research: Table of Specification

  2 pages
• RESERCH TITLE Master F4 An Evaluation On

  

  No ratings yet

  RESERCH TITLE Master F4 An Evaluation On

  26 pages
• The Background of The Study

  

  No ratings yet

  The Background of The Study

  13 pages
• A One-Month GMAT Study Plan

  

  100% (1)

  A One-Month GMAT Study Plan

  51 pages
• Research

  

  No ratings yet

  Research

  6 pages
• Page 175

  

  No ratings yet

  Page 175

  12 pages
• The Impact of CSR On Corporate Communication A Study of Central Danone Maroc

  

  No ratings yet

  The Impact of CSR On Corporate Communication A Study of Central Danone Maroc

  9 pages
• Behavior Assessment

  

  100% (1)

  Behavior Assessment

  33 pages
• Inquiry Vs Research

  

  No ratings yet

  Inquiry Vs Research

  12 pages
• Math 7

  

  No ratings yet

  Math 7

  19 pages
• Introduction To Qualitative Social Network Analysis

  

  No ratings yet

  Introduction To Qualitative Social Network Analysis

  8 pages
• HSC EES AT1 NOA 2024_25

  

  No ratings yet

  HSC EES AT1 NOA 2024_25

  8 pages
• AXIS Appraisal Tool

  

  No ratings yet

  AXIS Appraisal Tool

  8 pages
• M1112SP IVb 2

  

  No ratings yet

  M1112SP IVb 2

  4 pages
• Ep - 1

  

  No ratings yet

  Ep - 1

  7 pages
• Purpose of A Research Proposal-Mahi1

  

  No ratings yet

  Purpose of A Research Proposal-Mahi1

  5 pages
• Hots & Lots

  

  100% (1)

  Hots & Lots

  34 pages
• Ethics of Research

  

  No ratings yet

  Ethics of Research

  10 pages
• Theory of Personality Chapter-1

  

  0% (1)

  Theory of Personality Chapter-1

  13 pages
• Concepts in Research 1

  

  No ratings yet

  Concepts in Research 1

  50 pages
• Monitoring and Evaluation NOTES

  

  No ratings yet

  Monitoring and Evaluation NOTES

  38 pages
• Strategic Management Dissertation Proposal

  

  100% (2)

  Strategic Management Dissertation Proposal

  4 pages
• Review of Related Literature in Research Paper

  

  100% (2)

  Review of Related Literature in Research Paper

  4 pages
• Literature Review Consumer Attitude

  

  100% (2)

  Literature Review Consumer Attitude

  8 pages