Case-based research on democratization

ABSTRACT

Empirical research on democratization is dominated by case studies and small-N comparisons. This article is a first attempt to take stock of qualitative case-based research on democratization. It finds that most articles use methods implicitly rather than explicitly and are disconnected from the burgeoning literature on case-based methodology. This makes it difficult to summarize the substantive findings or to evaluate the contributions of the various approaches to our knowledge of democratic transition and consolidation. There is much to gain from a closer collaboration between methods experts and empirical researchers of democratization.

KEYWORDS:

• Methodology
• qualitative research
• case studies
• comparisons
• database
• process tracing

Introduction

What is the most common method to study democratization? Is it large-N statistical analysis, formal modelling, nested research designs, QCA, experiments, small-N qualitative comparisons, or single case studies?¹ It is safe to say that the majority of political scientists would answer: case studies. Coppedge writes about “a huge number of books and articles about the birth, death, or survival of democracy in dozens of countries (… ) more than any one scholar can digest.”² He even warns that “anyone unlucky enough to be buried under this small mountain [of single case studies and comparative histories] would surely be suffocated and crushed.”³ Whether case-based research provides a mountain to stand on or to be buried under is probably a matter of methodological preference, but two different sources attest to the popularity of case-based research in Comparative Politics in general and democratization studies in particular. Schedler and Mudde’s Dataset of Articles in Comparative Politics contains 581 empirical studies published in leading academic journals between 1989 and 2007.⁴ In these journals, quantitative studies outnumber qualitative studies, but among the latter, single case studies dominate, followed by small-N comparisons.⁵ Pelke and Friesen’s Democratization Articles Dataset, focused on democratization studies, is even more comprehensive, containing information on 3724 articles published between 1990 and 2016 in the leading journals in Comparative Politics and democratization research.⁶ Case studies make up almost half of the dataset, followed by qualitative comparisons.⁷

This article is a first attempt to take stock of qualitative case-based research on democratization. The methods literature tends to focus on exemplary studies, but we lack systematic evidence of what practitioners do.⁸ The analysis of case-based research on democratization is guided by two main questions. First, are these studies explicit about case selection and research design? Second, do they make use of the literature on case study research? The answers come from an analysis of a selection of articles in Pelke and Friesen’s dataset.⁹ For both case studies and small-N comparisons, the focus is on those journal articles that are most likely to exhibit methodological strengths, based on the coding in the dataset: the 31 qualitative studies coded as crucial, most-likely, and least-likely cases plus the 79 articles coded as using Mill’s methods of comparison.¹⁰ We will learn more from studies that make causal claims, justify case selection, and follow an established research design than we would from an in-depth examination of articles that merely aim to describe cases whose relevance is never explained.¹¹ The answers are mixed. On closer scrutiny, many studies are not explicit about case selection and research design, and references to the relevant methodological literature are scarce. However, purely descriptive articles are rare and most case-based research seeks to make a contribution to theory building.¹² The obvious lesson is that scholars are more likely to achieve their objectives when they are explicit about research design and case selection.

The article is organized in five sections. The next section provides an, of necessity, very brief overview of the methodological literature on case-based research. This prepares the ground for the following two empirical sections, which present a review of, first, case studies on democratization and, second, small-N comparisons on democratization. The conclusion formulates answers to the questions guiding this analysis and makes a plea for more engagement among methodologists and empirical students of democratization.

Case-based research

It is no exaggeration to speak of a boom in books on case-based research. Important new publications include Causal Case Study Methods,¹³ Designing Case Studies,¹⁴ Case Studies and Causal Inference,¹⁵ the second, revised edition of Case Study Research,¹⁶ and Process Tracing,¹⁷ which in many ways is a successor to the classic Case Studies and Theory Development in the Social Sciences.¹⁸ There was never a better time to find inspiration in the ingenious variety of qualitative, case-based research designs, to learn how to do case studies and small-N comparisons, and to learn how to do case-based research better.

One book has special relevance for students of democratization: Democratization and Research Methods.¹⁹ As is clear from the quote in the introduction, Coppedge has reservations about the widespread use of case studies in democratization research. Still, he admits that “many of the most influential theories in comparative politics originated in case studies.”²⁰ Coppedge defines a case as “a country observed during a period of time.”²¹ This definition works well for most research on democratization, which traditionally has been concerned with regime change at the national level in sovereign states.²² Why do a case study? Because of its conceptual validity, the possibility to derive new hypotheses, explore causal mechanisms, and model and assess complex relations.²³ Why not do a case study? Because of myopia, or the tendency to exaggerate the impact of short-term micro-causes; an inability to generalize; and an inability to distinguish causality from coincidence and spurious relationships, due to the lack of variation on causes and outcomes.²⁴

The classic literature in Comparative Politics recognizes three broad types of research using observational data: the statistical method (when the number of cases is large enough), the comparative method (small-N) and studies of single cases.²⁵ Instead, recent textbooks on case-based research cover both case studies and the comparative method.²⁶ This article follows recent textbooks in discussing case studies and small-N comparisons under the heading of “case-based research,” while it follows the early literature in discussing case studies separately from small-N comparisons.²⁷

Much of the contemporary methodological literature focuses on the relation between individual cases and the population from which they are drawn.²⁸ For Rohlfing, “the case is not interesting in itself (at least not in the first place), but for learning something about the population of cases from which it is drawn.”²⁹ Not everybody agrees. According to Blatter and Haverland, “investigators conducting applied research are often not interested in statistical generalization. They have an intrinsic interest in a specific case.”³⁰ Moreover, there is an alternative to case selection as sampling or the selection of cases to ensure representativeness vis-à-vis a population: cases can also be selected explicitly and purposefully for theory building. Rohlfing calls this “distribution-based” versus “theory-based” case selection strategies.³¹ As we shall see, the early literature on case studies in political science was less interested in generalizability and more in theory building.³²

There is no standard vocabulary to capture types of case studies. For example, the descriptive case study has been labelled “a-theoretical,” “configurative,” and “inductive.”³³ There seems to be more consensus on how to construct typologies of case studies. The basic breakdown is by research goal.³⁴ Is the main interest in the case itself or in its contribution to theory? Descriptive case studies and “interpretative” or “theory-guided” case studies both seek to enhance our knowledge of a particular case.³⁵ The first through description, the second through application of a theory to a particular case, without the intention to go beyond it. Theory-generating and theory-testing case studies, as the labels suggest, seek to contribute to theory building. Theory-testing case studies come in different forms. Eckstein describes the crucial case as one “that must closely fit a theory if one is to have confidence in the theory’s validity, or, conversely, must not fit equally well any rule contrary to that proposed.”³⁶ Most-likely and least-likely cases are variations on the crucial case. The requirement that a case must fit is replaced with a very high or low expectation that it will/should fit the theory. Crucial, most-likely, and least-likely cases play an important role in congruence analysis, described as “a small-N research design in which the researcher uses case studies to provide empirical evidence for the explanatory relevance or relative strength of one theoretical approach in comparison to other theoretical approaches.”³⁷ A case that is least-likely for the theory of interest to the author and most-likely for one or more alternative theories constitutes a “toughest test.”³⁸ Conversely, when a case is most-likely according to the theory of interest but least likely for others, then this is an “easiest test.”³⁹ Deviant case studies have several uses. They can “probe new explanations for Y, to disconfirm a deterministic argument, or to confirm an existing application (rare).”⁴⁰ For Lijphart, deviant cases studies are most helpful in uncovering additional variables that explain the outlier, thereby helping to modify the theory.⁴¹

Small-N comparisons are guided by Mill’s methods of agreement and difference, also known as the most different versus the most similar research design.⁴² More intuitive is De Meur and Berg-Schlosser’s terminology of “most similar systems with different outcomes” and “most different systems with the same outcome.”⁴³ The key assumption is that similarities cannot explain differences and vice versa. Options to compare a small number of cases do not end here. Gerring presents the diverse-cases method, which selects multiple cases in order to “represent” the full range of variation on the independent variable, the dependent variable, or their relationship.⁴⁴ Cases can be selected on the values of the key variables (for example high, middle, low) or to cover the main types in a typology.

Case studies in democratization research

According to Pelke and Friesen’s dataset, 907 out of the 1991 empirical studies on democratic transition and consolidation are single case studies (46%).⁴⁵ According to their coding, 85% of these case studies contain a causal claim and only 15% are purely descriptive. Less than a quarter of the studies explain why the case was selected and less than 1-in-10 make an attempt at generalization. For the following in-depth analysis of methodological practices in applied research, all 31 qualitative studies were selected that the dataset codes as crucial, most-likely, and least-likely cases. There are three reasons for focusing on crucial case studies. First, if “case studies are an incredibly powerful tool for examining whether concepts and theories travel, and whether (or not) they work in the same way in cases other than where they were originally developed,” then a focus on crucial case studies in democratization research will reveal more about these concepts and theories than other types of case study.⁴⁶ Second, crucial cases are selected in relation to theory, thereby providing a close connection between theory-building and empirical research on democratization.⁴⁷ Third, this category of case studies is arguably the most controversial, “with many scholars doubting the utility of case studies for hypothesis testing.”⁴⁸

The analysis is guided by five questions. First, what are the grounds for case selection according to the author(s) of the article? Second, is the article explicit about the type of case study/research design? Concretely, does it mention the terms “crucial,” “most-likely” or “least-likely”? Third, does the article refer to the methodological literature on case study research?⁴⁹ Fourth, does the article explicitly mention “process tracing” and refer to George and Bennett’s pioneering work? Process tracing has been called “perhaps the central within-case method,” inviting a check on its popularity in democratization research.⁵⁰ Fifth, does the article explain how close the case is to the theory? Rohfing complains that “most-likely and least-likely case studies are usually implemented in an informal manner” and he formulates three guidelines, which can be summed up as a responsibility to specify the probabilities of a confirming/disconfirming outcome.⁵¹ Two caveats are in order. First, this is a recent recommendation and therefore its application to the empirical literature will be mostly ex-post. Second, it may not be realistic to expect scholars “to list every empirical feature of the case and reflect on whether and how a given element influences the conditional likelihood.”⁵² However, it certainly does not seem too much to ask scholars to make explicit why they think a case is crucial and to estimate how likely or unlikely the outcome is in light of the theory to be tested, at least verbally if not actually stating probabilities.

The remainder of the section provides answers to all 5 questions, examining in turn the various types of case studies found among the 31 qualitative articles that Pelke and Friesen code as crucial cases. On closer scrutiny, there is only one explicit crucial case and one explicit most-likely/least likely case study. In addition, there are six deviant case studies and two typical cases. Table 1 lists all types of cases that could be identified. Each of these 13 studies is discussed in the text, grouped together under four headings: crucial cases, most-likely/least-likely cases, deviant cases, and typical cases.

Table 1. Case-based research on democratization: a breakdown.

Type of case in case studies	Type of small-N comparison
Crucial (1), “critical” (1), “significant” (1)	Explicit most-similar (2)
Most-likely (1), “best-case” (1)	Implicit most-similar (32)
Deviant (5), “exception” (1)	Explicit most-different (1)
Typical (1), “useful” (1)	Implicit most-different (7)
Methods references in case studies	Methods references in small-N comparisons
Eckstein 1975 (2)	George and Bennett 2005 (2)
Lijphart 1971 (1)	Przeworski and Teune 1970 (2)
George and Bennett 2005 (1)	Gerring 2007 (1)
Gerring 2007 (1)	Lijphart 1971 (1)
	Brady and Collier 2004 (1)

Source: See text.

Notes: The apostrophes indicate that a type of case is not part of standard typologies of case studies. Frequencies are given in parentheses.

Crucial cases

As can be seen in Table 1, of the 31 articles, only one is explicitly set-up as a crucial case study.⁵³ Jung selects Bosnia and Herzegovina as a crucial case to study the relationship between power sharing and democracy after the end of a civil war.⁵⁴ According to the author’s own coding of all the countries that ended their civil war through a negotiated settlement, Bosnia and Herzegovina has the largest extent of power sharing. This makes it “the most likely case for successful post-civil war democratization,” also because of the presence of international peace-keeping forces, another factor considered important for success.⁵⁵ Jung’s crucial case study is set up as a “rigorous test” of the power-sharing approach.⁵⁶ His aim is to highlight the dysfunctional features of Bosnia and Herzegovina’s institutions, blaming the Dayton Peace Agreement for most of the country’s current ills.⁵⁷

Taylor’s study of Russia as a case of federalism in a hybrid regime, which combines elements of democracy and autocracy, is modelled as a “plausibility probe” of a hypothesis derived from the literature on democratic federations.⁵⁸ The notion of the plausibility probe was invented by Eckstein as an attempt to “determine whether potential validity may reasonably be considered great enough to warrant the pains and costs of testing.”⁵⁹ Case studies are eminently suited for this purpose. Eckstein does not specify how cases should be selected in plausibility probes, though in his own study Norway is presented as “critical” for the theory he wants to explore.⁶⁰ Taylor selects Russia as a “significant case,” significant because of its real-world importance and because changes in relations between the centre and the subunits over time offer the researcher relevant variation to work with.⁶¹ In the end, though, the study reads more like a theory-guided case study then a plausibility probe. There is no reflection on how the Russian case relates to other hybrid federations, and nothing is said about future research that would develop the plausibility probe into a proper test.

Törnquist evaluates the explanatory power of three theories to explain peace and democracy in Aceh: the liberal argument, sequencing, and transformation. For each, he formulates expectations. Aceh is presented as a “critical” case for these theories.⁶² Törnquist thus combines a congruence test with a crucial case study. The evidence is mixed, however, as “none of the general arguments are completely refuted or confirmed in the context of Aceh.”⁶³ What that means for the explanatory power of the three theories is left unexplored.

Most-likely/least-likely cases

One other study in this set explicitly adopts a most-likely/least-likely research design. Haddad argues that for some theories, especially those centred on socio-economic development and institutions, Japan is a most-likely, even “almost inevitable,” case of democratization whereas for other theories, for example on political culture, the country is a least-likely or even “impossible” case.⁶⁴ Haddad’s interest, though, is not in testing any of these theories, but in developing a new one. She offers a “theory-building case study,” focused on the relationship between civil society and the state in Japan.⁶⁵ Rakner’s “best-case scenario” is best understood as implicit “most-likely” case selection. She examines Zambia, a country that failed to consolidate democracy despite a promising beginning.⁶⁶ Her explanation centres on the lack of party and party system institutionalization and on international aid.

Deviant cases

Pelke and Friesen’s list of crucial cases in democratization research also contains what turns out to be a cluster of deviant cases, all contributions to a special issue in this journal.⁶⁷ Following a mixed-methods design, the editors first selected deviant cases based on a summary of the state-of-the-art on the drivers of democratization, often large-N studies, then present five case studies, and conclude with lessons learned.⁶⁸ The case studies of Benin,⁶⁹ Botswana,⁷⁰ Costa Rica,⁷¹ India,⁷² and Mongolia,⁷³ are “geared towards detecting important variables that can explain the unexpected transition to and consolidation of democracy.”⁷⁴ The conclusion reinforces the impression that the goal is not theory testing, but the generation of new hypotheses.⁷⁵ McMillan’s analysis of India provides a good illustration of how a deviant case study can help to modify existing theory, in this case modernization theory. He identifies two “main crucial factors” that help explain India’s transition to democracy: the ideology and organization of the Indian nationalist movement and the administrative and representative framework established by the British rulers.⁷⁶

In contrast, Meyns’ analysis of the “exception” of Cape Verde misses the opportunity to systematically investigate what makes the country a rare success story of democratic reform in Africa.⁷⁷ There are no theories and no explicit hypothesis, although it is possible to recognize one in the following explanation of peaceful transformation in Cape Verde, which focuses on the behaviour of political actors: “Drawing on intellectual traditions of debate within their society and aware of the vulnerability of their country, they have developed a nonviolent political culture that has shaped the process of democratic transition.”⁷⁸ If this is the conclusion, the case study could be considered hypothesis generating. However, the conclusion would have been much stronger if it had been the outcome of a deviant case study ruling out alternative explanations, arriving at political culture as the key factor.

Typical cases

Other articles appear to be typical case studies, meaning that cases are selected to represent a broader set for which they are representative. Faust et al. deliberately select Zambia as a “typical case” of Multi-Donor Budget Support.⁷⁹ This allows the authors to draw generalizing inferences with, moreover, clear policy implications. However, it is left to the reader to judge how well the conclusions based on the case of Zambia travel to similar cases. Le Van’s focus on Nigeria as a “useful case” likewise is motivated by the desire to use information from a typical case to make a more general statement about the relevance of Tocqueville’s insights on civil society for understanding contemporary processes of democratization in Africa.⁸⁰ Throughout the article, it is assumed rather than demonstrated that what happens in Nigeria is in many ways representative for the rest of Africa. In other words, both studies would have been stronger if they had paid more attention to case selection.

Unfortunately, for most studies coded as crucial cases in the Pelke and Friesen dataset it is not possible to ascertain what kind of case study the author(s) had in mind when selecting their cases. Arguably, this hinders authors in making the most of their findings. For example, Hughes concludes her analysis of parties, protest, and pluralism in Cambodia with the statement that “crucially, for democratization, the development of urban protest as a major form of political activism has eclipsed military confrontation as a political strategy for parties in Cambodia.”⁸¹ Hughes highlights the broader relevance of her findings, which demonstrate that “the pessimistic view of political parties in the South gives insufficient weight to the positive importance of political parties.”⁸² This conclusion would have been even stronger if it had been the outcome of a more explicitly theory-oriented research design using Cambodia as a deviant case to qualify and modify expectations from the 1990s literature on the democratizing potential of political parties.

Starting with the 31 qualitative articles that Pelke and Friesen coded as crucial case studies, this section has reviewed all 13 articles where case selection could be identified. Only one of these is an explicit crucial case and only one an explicit most likely/least likely case. Positively, both articles elaborate on the expected probabilities, as recommended in the literature on case-based research. In addition, we find six deviant case studies and two typical cases. The other studies do not appear to follow a particular type of case study. This lack of methodological self-consciousness arguably weakens the authors’ findings and limits their theoretical relevance. Only four empirical studies refer to the methodological literature on case studies in political science, suggesting a disconnect. Only one of the articles listed as a crucial case study in Pelke and Friesen’s dataset engages in process tracing, despite the popularity of this method of within-case analysis.⁸³ Finally, one-third of the case studies makes an attempt at generalization. Using Rohlfing’s distinction between “distribution-based” and “theory-based” case selection strategies, it is clear that most students of democratization have opted for the latter. While legitimate, this leave readers with questions about the generalizability of the findings. More attention to a theory’s scope conditions would help. Finally, the collection of articles on deviant cases of democratization shows the potential of case studies to contribute to theory building.

The comparative method in democratization research

Pelke and Friesen list 418 empirical “comparative case studies” on democratic transition and consolidation. Almost all (93%) make causal claims, but only 248 studies (59%) explain why the sample/cases were chosen and 83 (20%) are coded as explicitly following Mill’s methods. This subset of studies is selected here as the most promising starting point for an examination of the way in which the comparative method is used in small-N research on democratization. After all, this selection of journal articles seems to most closely follow good practice in case selection and research design. The analysis is guided by four questions. First, what are the grounds for case selection according to the author(s) of the article? Second, is the article explicit about the type of comparison? Concretely, does it mention the terms “most similar,” “most different,” “method of agreement” or “method of difference”? Third, does the article refer to the methodological literature on case-based research? Fourth, does the article explicitly mention “process tracing” and refer to George and Bennett’s pioneering work?

A closer look at the articles that Pelke and Friesen’s dataset codes as following Mill confirms that the majority indeed do so. Mill’s method of difference is clearly more popular than the method of agreement.⁸⁴ As can be seen in Table 1, direct references to either Mill’s work or explicit mention of his methods, though, are rare. Only 3 of the 42 articles that select cases based on similarities or differences judged to be theoretically relevant are explicit about the kind of comparison they make. Most often, authors paraphrase Mill’s logic, listing the similarities among their cases as justification for case selection. For example, already in the opening sentence of his paired comparison, Renwick stresses how

the transitions from communism in Hungary and Poland in 1989 shared many features: both occurred following years of tentative liberalization; both were pacted, proceeding through round-table negotiations between regime and opposition elites; and both were in significant part initiated by regime reformers.⁸⁵

He then focuses on the differences to explain the divergent trajectories of these countries. Another example from the 32 implicit most-similar research designs is Carbone’s comparison of health policy in Ghana and Cameroon. His aim is to show that democracy matters for people’s lives, focusing on health care. “Ghana and Cameroon make a good choice for comparison since they are reasonably similar countries, except that, from the early 1990s, their political trajectories followed dramatically divergent paths,” with Ghana democratizing and Cameroon remaining authoritarian despite the introduction of multi-party elections.⁸⁶

Eight articles in the dataset follow a most-different research design (or Mill’s method of agreement). Only one is explicit: Field and Siavelis’s study of the way political parties select candidates for founding elections. They first argue why qualitative case studies are an appropriate method for their purposes. The authors then explain why they opt for a most-different research design and chose Chile and Spain, citing relevant literature on social science methods. This comparison allows them to link, theoretically and empirically, a particular type of candidate selection strategy, common among the parties in both countries, to regime uncertainty, another thing they have in common.⁸⁷ More often, though, most-different research designs are implicit, as in Hill’s justification of his case selection:

This paper has chosen to compare mission strategies in two states that are very different: Bosnia and Afghanistan. Given the enormous differences between these two states, if the respective democracy promotion strategies were developed on the basis of analysis of local conditions (i.e. “case-by-case”) then the policies implemented could be expected to be very different in each case.⁸⁸

One issue with the implicit use of the method of agreement and difference concerns the theoretical status of the similarities and differences. The selected countries might be very similar in some respects, but how important are these similarities theoretically? Does case selection control for alternative explanations? Are the similarities and differences, even if real, directly relevant to the causal model? Being explicit about the adoption of Mill’s methods might encourage a more rigorous specification of the theoretical relevance of similarities and differences. That said, there are no guarantees. In one of only two explicit most-similar research designs, Green, referring to Gerring, compares the impact of decentralization on opposition parties in Sudan and Ethiopia.⁸⁹ These countries are chosen because as

two of Africa’s largest and most populated countries, these two neighbouring states present very comparable case studies for an examination into the relationship between decentralization and opposition politics. Historically both countries have suffered from numerous civil wars, famines and violent regime changes, and both border other countries which have suffered from similar problems. Thus they are ideal for a “most similar” case study comparison.⁹⁰

Whether these cases do indeed make an ideal pairing depends, of course, on how relevant the presence of civil war and famine is for the fate of opposition party unity.⁹¹

Several studies deliberately incorporate cases with a positive and negative outcome. A favourite comparison is that between successful democratization in Tunisia and initially more difficult and ultimately failed democratization in Egypt.⁹² The challenge is to isolate the factors that made the difference. This reflects a more general problem with Mill’s methods as tools of causal identification and control of alternative explanations. Beach and Pedersen even argue that small-N comparisons “cannot stand alone when making confirming causal inferences. Instead, they should always be coupled with within-case studies.”⁹³ Process tracing can aid this task, but Brumberg is actually critical of what he calls “reading history backward,” which tricks the scholar to “trace any particular turn of events during a transition to its apparently sufficient causation in some historical legacy or point of origin.”⁹⁴ In fact, explicit process tracing is rarely used in the studies reviewed here. Hale presents a “process-tracing paired comparison” of regime change in Ukraine and Kyrgyzstan.⁹⁵ The article traces the impact of what Hale terms, borrowing language from the experimental methods, a “crucial ‘treatment’” in the form of a different institutionalization of power sharing.⁹⁶ The interest is explicitly in the impact of the “treatment,” not its origin. Buckley selects Ireland and Senegal because they are “paradigmatic tough cases”: in both countries democracy was likely to succumb to the dynamics between religion and secularism.⁹⁷ However, in both countries, though in the case of Senegal with some delay, democracy prevailed. Buckley explains this outcome through a focus on the way in which institutions shape actor preferences. His interest is more in theory building than the cases themselves, as is clear from the statement that “the empirics (…) draw on two states, but the basic strategy for inference is within-case process tracing of the hypothesized effects of institutional structure.”⁹⁸

Some studies follow an implicit diverse cases selection strategy. Heper compares Germany and Turkey to “demonstrate that different degrees of stateness have significantly different consequences for the consolidation of democracy.”⁹⁹ The Ottoman-Turkish and Prussian-German cases are selected as “polar types.”¹⁰⁰ Collier is interested in the relationship between labour policies and regime change in Latin America in the 1940s. After discussing the limitations of Mill’s methods she decides to examine two perspectives on this relationship, one stressing internal trajectories and the other external influences, in combination. To do so, Collier selects four “polar types” (Brazil, Chile, Mexico, Venezuela), though she does not elaborate on the case selection, merely promising that it “will become evident” in the analysis.¹⁰¹

Villalón shows how one can do a deviant case study with multiple cases. He zooms in on the new democracies in Mali, Niger, and Senegal, noting that “neither prevailing theories of democracy nor those of politics in Muslim societies prove particularly useful in accounting for these exceptional cases on the southwestern edge of the Muslim world.”¹⁰² His interest is in the way democracy and religion interact and shape each other. The conclusion, based to a large extent on field research, may come as a surprise: in all three West African countries “the role of religion in the public sphere has significantly expanded. This public presence has, in turn, produced a ‘democratization’ of the religious sphere.”¹⁰³

Sandbakken shows how one can do congruence analysis with multiple cases.¹⁰⁴ She compares three African rentier states (Algeria, Libya and Nigeria) that have little in common except their dependence on oil. The resource curse assumes a negative relationship between oil wealth and democratization and that is indeed what Sandbakken finds. However, her research design is more ambitious than a simple confirmation of an established theory: Sandbakken’s main interest is in identifying the casual mechanism and to that end she formulates and tests three hypotheses explaining the link between resource wealth and lack of democracy. In other words, Sandbakken tests three variants of the same theory with three African cases. She finds that “overall, the cases show that rentier state theory does apply to African rentier states (…) However, different parts of the theory suit better to some of the cases than others.”¹⁰⁵

In sum, Mill’s methods inform comparative research into democratization, though mostly implicitly. References to the literature on case-based research are rare. This is not for lack of ambition: almost all studies aim at theory building and half of them make an explicit attempt at generalization. In order for small-N comparisons to realize their full potential, they should be explicit about the logic of comparison and case selection, supporting their choices with reference to the relevant literature in political science methodology.

Conclusion

Writing 20 years ago, Peters noted that “case studies are often conducted poorly.”¹⁰⁶ Is that still true today? The analysis of case-based research on democratization in this article was guided by two questions. First, are these studies explicit about case selection and research design? Second, do they make use of the literature on case-based research? The answers to both questions are mostly negative: the overwhelming majority of studies evaluated here are not explicit about case selection and research design, do not refer to the relevant methodological literature, and make little use of explicit process tracing as a method of causal inference.¹⁰⁷ These conclusions apply to both case studies and small-N comparisons of democratic transition and consolidation. Since these articles were selected because they seemed to correspond most closely to good practice, it is likely that the weaknesses identified here can be found in the broader literature.¹⁰⁸ These studies often exemplify dedicated field research, deep knowledge of cases, precise conceptualization, weighing of different explanations, helpful description, and genuine insights. Purely descriptive articles are rare and most case-based research seeks to make a contribution to theory building. This leads to a conclusion that should be uncontroversial and that is not limited to research on democratization: scholars engaged in case studies or small-N comparisons are more likely to achieve their research objectives when they are explicit about their research design and case selection. Consultation of the literature on research methodology can help with this task. However, like Beck’s review of comparative research on revolutions, this study of case-based research on democratization reveals a “disconnect between methodologists and practitioners.”¹⁰⁹ The literature on democratization rarely refers to the methodological literature on case-based research, whether classic or recent. Is this because “transitology” is an art more than a science, as Schmitter famously suggested?¹¹⁰ More likely, these bodies of literature have developed side by side, with only few scholars combining the two.¹¹¹

Because research design and case election are often not explicit, it is difficult to summarize the substantive findings of case-based research on democratization.¹¹² It is impossible to evaluate the usefulness of Mill’s methods for the study of democratization if even scholars who write about their cases as being very similar or very different do so without grounding their approach explicitly and systematically in the comparative method. It should be the exception, not the rule, to encounter a case study that does not ask the basic question what kind of case study it is.¹¹³ Case-based research on democratization is surprisingly strong in causal analysis and theory building, but lacks an interest in generalization. Although some see this as a problem inherent in the method, explicit attention to the theory’s scope conditions can aid scholars in going beyond their case(s). Does the lack of crucial case studies in democratization research reflect the limits of this particular method in testing theories of democratic transition and consolidation, the poor state of theorizing in this field, or is it simply an opportunity to be seized? If the experience with deviant cases is any guide, the answer would have to be the latter. The collection of well-designed deviant case studies discussed here shows how case studies can contribute to theory building in democratization research.¹¹⁴

Beyond increased methodological self-consciousness, the methods literature on case-based research has at least three things to offer students of democratization: it can be a source of inspiration (what to do), a guide (how to do it), and a benchmark (how to do it better). As a source of inspiration, the literature on case-based research methods provides scholars with an array of options. There are many types of case studies to choose from and the comparative method has more to offer than Mill’s two classic options. As a guide, the methods literature provides rules for case selection and research design. Beach and Pedersen conclude many of their chapters with a set of guidelines that read like a manual.¹¹⁵ As a benchmark, the methods literature, through the discussion of best practices and the formulation of guidelines, can help students of democratization to evaluate and improve their research. It is a welcome development that the methods literature provides ever more explicit guidelines. Bennett and Checkel’s list with 10 best practices for process tracing is an excellent example.¹¹⁶ In other words, there is much to gain from a closer collaboration between methods experts and empirical researchers of democratization.¹¹⁷

Acknowledgements

The author is heavily indebted to the editors of this special issue, Michael Coppedge and David Kuhn, for their guidance and support. Thanks are also due to the two anonymous reviewers, who pushed me to clarify the argument and improve the presentation of the data.

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Notes on contributors

Matthijs Bogaards

Matthijs Bogaards is a visiting professor in the Department of Political Science at the Central European University in Budapest, Hungary. He has published widely on comparative politics, with a special interest in democracy, democratization, and hybrid regimes.

Notes

1. For a comprehensive overview of the manifold methods used in the study of democratization, see the other contributions to this special issue.

2. Coppedge, Democratization and Research Methods, 115. Similarly, Geddes concludes that scholars have “amassed an astonishing amount of ‘data,’ mostly in the form of case studies,” whereas “our theoretical understanding remains thin.” Geddes, What Do We Know About Democratization, 142–3.

3. Ibid.

4. Schedler and Mudde, Data Usage.

5. In this article, the term case studies always refers to studies of single cases.

6. These journals are: Journal of Democracy, Democratization, Comparative Politics, Comparative Political Studies, and World Politics. See Pelke and Friesen, Democratization Articles Dataset.

7. Pelke and Friesen, Democratization Articles Dataset, table 3.

8. Beck, The Comparative Method, 536.

9. The author thanks Paul Friesen and Lars Pelke for kindly sharing the data.

10. For the purpose of this article, Pelke and Friesen’s dataset is thus primarily a starting point, used for identifying relevant articles. Information about topic, research design, causal claims, and generalization comes from Pelke and Friesen’s data. All other information is the responsibility of the author of this article.

11. The analysis here includes articles that the dataset codes as having empirical content and with democratic transition or consolidation as their primary theme. These criteria follow the classic preoccupation of the democratization literature with processes of democratic transition and consolidation, priorities that are also reflected in the numbers: two-thirds of all articles in the dataset deal with democratic transition and consolidation. Pelke and Friesen, Democratization Articles Dataset, table 4.

12. See Pelke and Friesen, Democratization Articles Dataset.

13. Beach and Pedersen, Causal Case Study Methods.

14. Blatter and Haverland, Designing Case Studies.

15. Rohlfing, Case Studies and Causal Inference.

16. Gerring, Case Study Research. The first edition is from 2007.

17. Bennett and Checkel, Process Tracing.

18. George and Bennett, Case Studies and Theory Development.

19. Coppedge, Democratization and Research Methods.

20. George and Bennett, Case Studies and Theory Development, 31.

21. Coppedge, Democratization and Research Methods, 116.

22. Ragin’s, What is a Case?, notion of “casing” can be understood as the research process that turns an empirical unit into a theoretically relevant case.

23. George and Bennett, Case Studies and Theory Development, 19–22.

24. Ibid., 120–8.

25. Lijphart’s suggestion that theory-developing case studies are “implicitly comparative,” though probably a compliment, weakens their methodological distinctiveness, Comparative Politics, 692–3.

26. Rohlfing, Case Studies and Causal Inference; Blatter and Haverland, Designing Case Studies; and Beach and Pedersen, Causal Case Study Methods. Halperin and Heath, Political Research, cover case studies in a chapter on “comparative research.”

27. Another common distinction is that between case-oriented and variable-oriented research. See Della Porta, Comparative Analysis.

28. The shift seems to coincide with King et al.’s advocacy of the logic of quantitative inference in qualitative research. King, Keohane, and Verba, Designing Social Inquiry; McKeown, Case Studies and the Limits of the Quantitative Worldview.

29. Rohlfing, Case Studies and Causal Inference, 61. Similarly, Gerring: “The case(s) identified for intensive study is chosen from a population and the reasons for this choice hinge upon the way in which it is situated within that population.” Gerring, What is a Case Study, 646.

30. Blatter and Haverland, Designing Case Studies, 69.

31. Rohlfing, Case Studies and Causal Inference, 62.

32. Lijphart rejected outright the possibility of generalizing on the basis of a single case. Lijphart, Comparative Politics, 691.

33. In, respectively, Lijphart, Comparative Politics; Eckstein, Case Study and Theory; and Levy, Case Studies.

34. Lijphart, Comparative Politics; Levy, Case Studies; and Rohlfing, Case Studies and Causal Inference.

35. The first term is from Lijphart, Comparative Politics, the second from Levy, Case Studies.

36. Eckstein, Case Study and Theory, 118, emphasis in original. For Eckstein, extreme cases are candidates for investigation as a crucial case; they do not constitute a category on their own.

37. Blatter and Haverland, Designing Case Studies, 144.

38. George and Bennett, Case Studies and Theory Development, 121.

39. Ibid., 122.

40. Seawright and Gerring, Case Selection Techniques, 297.

41. Lijphart, Comparative Politics, 692.

42. Przeworski and Teune, The Logic of Comparative Social Inquiry.

43. De Meur and Berg-Schlosser, Comparing Political Systems.

44. Gerring, Case Selection, 650.

45. Pelke and Friesen, Democratization Articles Dataset, table 4.

46. Halperin and Heath, Political Research, 207.

47. See Blatter and Haverland, Designing Case Studies, 175–8.

48. Levy, Case Studies, 6.

49. For a similar approach, see Beck, The Comparative Method. Beck examined all journal articles that use comparative analysis to explain revolutions.

50. Bennett and Checkel, Process Tracing, 4. They define process tracing as “the analysis of evidence on processes, sequences, and conjunctures of events within a case for the purpose of either developing or testing hypotheses about causal mechanisms that might causally explain the case,” 7.

51. Rohfing, Case Studies and Causal Inference, 85.

52. Ibid., 88.

53. The abstract of Mantilla’s article labels Mexico a crucial case, but this is not repeated or elaborated upon in the text, Democratization and the Secularization of Religious Parties. Despite recognizing Prussia as a crucial case for theories of the role of landed elites and political regimes, Ziblatt does not opt for a (qualitative) crucial case design, but instead prefers a (quantitative) within-case comparison based on electoral districts. The number of observations allows for statistical analysis, making Ziblatt’s case study one of the surprisingly frequent quantitative case studies identified by Schedler and Mudde. Ziblatt, Does Landholding Inequality Block Democratization? Schedler and Mudde, Data Usage.

54. Jung, Power Sharing and Democracy Promotion.

55. Ibid., 492.

56. Ibid., 493.

57. This conclusion depends on the author’s use of the Polity dataset, which continues to classify Bosnia and Herzegovina as a case of “foreign interruption,” whereas for Freedom House the country was an electoral democracy between 1996 and 1998 and again since 2008. For a discussion of these measures of democracy, see Bogaards, Where to Draw the Line?

58. Taylor, Force and Federalism, 421.

59. Eckstein, Case Study and Theory, 108.

60. Ibid., 111.

61. Taylor, Force and Federalism, 422.

62. Törnquist, Dynamics of Peace and Democratization, 826.

63. Ibid., 840.

64. Haddad, The State-in-Society Approach, 1002, 1001.

65. Ibid., 998.

66. Rakner, Institutionalizing the Pro-Democracy Movements, 1118.

67. Special issue on Deviant Democracies, Democratization 15, no.4 (2008), edited by Renske Doorenspleet, Petr Kopecký and Cas Mudde.

68. For a similar approach, see Seeberg, Mapping Deviant Democracy.

69. Gisselquist, Democratic Transition.

70. Good and Taylor, Botswana.

71. Booth, Democratic Development.

72. McMillan, Deviant Democratization.

73. Fritz, Mongolia.

74. Doorenspleet and Kopecký, Against the Odds, 709.

75. Doorenspleet and Mudde, Upping the Odds, 816.

76. McMillan, Deviant Democratization, 741.

77. Meyns, Cape Verde.

78. Ibid., 164.

79. Faust, Leiderer, and Schmitt, Financing Poverty Alleviation, 445.

80. Le Van, Questioning Tocqueville in Africa.

81. Hughes, Parties, Protest and Pluralism, 183.

82. Ibid.

83. Mantilla, Democratization and the Secularization of Religious Parties.

84. This might be for good reason. According to Ragin, Mill’s method of agreement is “generally regarded as an inferior technique.” The Comparative Method, 36.

85. Renwick, Why Hungary and Poland Differed, 36.

86. Carbone, Do New Democracies Deliver Social Welfare?, 175.

87. The exceptional clarity of their research design is undermined by the statement in the conclusion that their research should be seen as a “hypothesis generating case study.” Field and Siavelis, Endogenizing Legislative Candidate Selection Procedures, 816.

88. Hill, Exploring USAID’s Democracy Promotion, 1000.

89. Green, Decentralisation and Political Opposition. The other explicit most-similar research design in Pelke and Friesen’s dataset is Ingram, Crafting Courts in New Democracies, a comparison of judicial reform in three Mexican states.

90. Green, Decentralisation and Political Opposition, 1088.

91. For an overview of factors affecting opposition unity in Africa, see Bogaards, Electoral Alliances.

92. Brumberg, Transforming; Stepan and Linz, Democratization Theory and the “Arab Spring.”

93. Beach and Pedersen, Causal Case Study Methods, 239.

94. Brumberg, Transforming, 100.

95. Hale, Formal Constitutions in Informal Politics, 588.

96. Ibid., 589.

97. Buckley, Beyond the Secularization Trap, 448.

98. Ibid., 447.

99. Heper, The Strong State, 169.

100. Ibid., 170.

101. Berins Collier, Combining Alternative Perspectives, 2.

102. Villalón, From Argument to Negotiation, 375.

103. Ibid., 388.

104. Sandbakken, The Limits to Democracy.

105. Ibid., 150.

106. Peters, Comparative Politics, 137.

107. These problems are not unique to research on democratization. Writing about case studies in international relations, Maoz writes that “unfortunately” advances in the methods literature “have gone largely unnoticed by most practitioners.” Maoz, Case Study Methodology in International Studies, 471.

108. The selection criteria applied in this article capture half of all articles coded as following Mill’s methods and one third of all articles coded as crucial cases in the Democratization Articles Dataset. See Pelke and Friesen, Democratization Articles Dataset, table 6.

109. Beck, The Comparative Method in Practice, 551.

110. Schmitter, Transitology.

111. Coppedge, Democratization and Research Methods, is an exception. See also Landman, Issues and Methods, 185–215.

112. Even Beck, The Comparative Method in Practice, who examines all comparative qualitative studies on revolutions published between 1970 and 2009, makes no attempt to summarize the substantive findings.

113. “Transparency” is the first of four general principles for case studies formulated by Maoz, Case Study Methodology in International Studies, 471. See also Gerring and Cojocaru, Selecting Cases for Intensive Analysis, who finish their overview of types of case studies with “a plea for transparency,” 417–18.

114. For a forceful plea for deviant case studies, see Seawright, The Case for Selecting Cases That Are Deviant.

115. Beach and Pedersen, Causal Case Study Methods.

116. Bennett and Checkel, Process Tracing.

117. While this conclusion is made on the basis of a review of studies on democratization, it obviously has broader relevance.