Introduction
Should scholars engaged in comparative research on democracy treat the distinction between democracy and nondemocracy as a dichotomy, or in terms of gradations? This recurring and much debated question has important implications for how research is organized, for how data are collected and analyzed, and for inferences about the causes and consequences of democracy. It also serves as a reminder that discussions of research design in political science must pay central attention to conceptual issues.
Among the authors who have advocated an approach based on grading and ranking, Bollen and Jackman argue that “democracy is always a matter of degree” and that treating it as dichotomous is a “flawed” practice (Reference Bollen and Jackman1989: 618, 612). A graded perspective is likewise adopted by Dahl, using the term polyarchy (Reference Dahl1971: 2, 8, 231–35; Reference Dahl1989: 241, 316–17), and later by Coppedge and Reinicke (Reference Coppedge and Reinicke1990). By contrast, Sartori finds that treating the distinction between democracy and nondemocracy in graded terms is an analytically “stultifying” exercise in “degreeism,” which misses the basic fact that political systems are “bounded wholes” (Reference Sartori1987: 184; also Reference Sartori1991: 248). Other scholars who have adopted a dichotomous approach include Linz (Reference Linz, Greenstein and Polsby1975: 184–85), Huntington (Reference Huntington1991: 11–12), and Geddes (Reference Geddes1999). Przeworski and collaborators have specifically rejected Bollen and Jackman’s argument as “confused” because it does not recognize that regimes “cannot be half-democratic: there is a natural zero point” (Alvarez et al. Reference Alvarez, Cheibub, Limongi and Przeworski1996: 21). Their position is especially striking because their larger project (Przeworski et al. Reference Przeworski, Alvarez, Cheibub and Limongi1996; Przeworski and Limongi Reference Przeworski and Limongi1997) is based on quantitative data and sophisticated forms of statistical analysis. Yet when it comes to measuring democracy versus nondemocracy, they select a dichotomy.
We see an interesting puzzle here. The choice of a dichotomy in effect places this distinction at what is traditionally viewed as the lowest level of measurement (Stevens Reference Stevens1946; Roberts Reference Roberts1976: 492–93). This choice thereby appears to underutilize more fine-grained information that may routinely be available about differences among regimes. Yet both Sartori and Przeworski and collaborators are convinced that this lowest level of measurement is more valid in conceptual terms. This puzzle points to a question: What, indeed, are the grounds for viewing this as a valid dichotomy, and not a “false dichotomy”?
This is an important question. First, quite apart from the scholars who are explicitly debating this choice, large numbers of qualitative researchers (on the one side) and quantitative researchers (on the other side) in effect take a position on this issue without ever directly addressing it. Second, this choice is important because it affects substantive findings of research on democracy. Although alternative dichotomous and graded measures are strongly correlated with one another (Alvarez et al. Reference Alvarez, Cheibub, Limongi and Przeworski1996: 21), Elkins (Reference Elkins1999) has shown that, in assessing the impact of regime type on the initiation of war, a graded measure reveals interesting incremental effects that would not be detected with a dichotomy. He likewise shows that, in studies of the effect of regime type on political stability, the use of a single cut-point can mask a relationship that emerges only if one looks at a different cut-point. Similarly, Coppedge (Reference Coppedge and Midlarsky1997: 181, 189–97) finds that in cross-national tests evaluating explanations for levels of polyarchy, different results emerge depending on which cut-point is employed in creating a dichotomous measure of polyarchy. Restricting the analysis to any single dichotomous cut-point can thus obscure potential findings.
Given that the findings of research can be influenced by these choices, it is important to examine the conceptual reasoning that justifies alternative approaches. However, somewhat surprisingly, recent methodological writing in political science provides insufficient guidance for dealing with such questions, and various contributors to recent debates on research design have called for greater attention to conceptual issues (Brady Reference Brady1995: 16–18; Collier Reference Collier1995: 463; Laitin Reference Laitin1995: 455–56; Munck Reference Munck1999).
This chapter examines the conceptual justifications that lead scholars to choose a dichotomous or graded approach. The first section, “Concept Formation and the Burden of Demonstration,” focuses on general methodological arguments about concept formation, dichotomies, and gradations. The second section, “Examples of Generic Justifications,” reviews examples of the generic justifications employed by prominent authors in the literature on democracy. The third section, “Toward More Specific Justifications,” considers more specific justifications, which we believe provide a better rationale for the choice between dichotomies and gradations.
In favoring these more specific justifications, we adopt a pragmatic position.Footnote 1 While recognizing that usage is shaped and constrained by the broader scholarly understanding of a concept’s meaning, we hold that specific methodological choices are often best understood and justified in light of the theoretical framework, analytic goals, and context of research involved in any particular study. As theory, goals, and context evolve, choices about concepts likewise may evolve.
To explicate our pragmatic approach, it is useful to identify two interrelated priorities that underlie this perspective. First, it rejects the idea that there is a single correct, or “best,” meaning for all concepts and views the search for a single best meaning as frequently being an unproductive enterprise. Second, this approach focuses on understanding how alternative meanings are connected with the specific goals and context of research. Thus, how scholars understand and operationalize a concept can and should depend in part on what they are going to do with it.
Our approach thus shares important concerns with the tradition of concept analysis that, in conjunction with a broad focus on the structure of meaning explores variations in concept usage among different authors and different schools of thought. This tradition is identified both with the field of political theory (Gallie Reference Gallie1956; Pitkin Reference Pitkin1967; Freeden Reference Freeden1996), and also with studies that draw directly or indirectly on Sartori’s (Reference Sartori and Sartori1984b: 40–42) methodology for the reconstruction of concepts. Examples of this latter approach include studies of consensus (Graham Reference Graham and Sartori1984), elite (Zannoni Reference Zannoni1978), ideology (Gerring Reference Gerring1997), political culture (Patrick Reference Patrick and Sartori1984), revolution (Kotowski Reference Kotowski and Sartori1984), and social movement (Diani Reference Diani1992). This tradition generally avoids preemptively ruling out particular meanings or usages, and instead focuses on understanding each usage in its own terms.
Our pragmatic approach also shares important concerns with the focus on “construct validity,” which is one central consideration in efforts by quantitative researchers in political science to evaluate choices about concepts and operationalization (Zeller and Carmines Reference Zeller and Carmines1980: 79–81). A central goal in assessments of construct validity is to evaluate whether a given operationalization of a concept, when used in testing a well-established hypothesis, yields results that are plausible and interesting in light of theoretical expectations regarding that hypothesis. With regard to the concept of democracy, Elkins (Reference Elkins1999) is an excellent example of this approach.
We share with the construct validity approach a concern with the details of how particular concepts are actually used in exploring specific research questions. However, in contrast to this approach, we address this concern in relation to a broader range of issues about how concepts are applied and understood in empirical research. For example, we consider the implications of treating democratization as a well-bounded “event” and the conceptual requirements for analyzing subtypes of democracy, as well as issues of the empirical distribution of cases and normative assessment that arise in specific contexts of research.
These questions of justifying choices about concepts are complex, and we wish to underscore two issues not addressed here. First, we take as given the procedural definition of democracy that has predominated in the recent comparative literature on democratization. The interesting question of dichotomies in relation to other definitions of democracy is not addressed. Second, if a scholar adopts gradations, a further choice concerns the choice of procedures for aggregating observations in scales.Footnote 2 This choice is crucial but is likewise beyond our focus.
Concept Formation and the Burden of Demonstration
This section explores two claims about dichotomies: (1) they are fundamental because concept formation is inherently based on classificatory reasoning; and (2) they can be justified through arguments about object concepts and bounded wholes. We conclude this section by considering where the burden of demonstration should lie in justifying a dichotomous or graded approach. For more than three decades, Sartori has been the scholar most centrally concerned with these issues.Footnote 3 Focusing on his views offers a productive way of exploring this debate, particularly because his position continues to be important in research on concept formation and on democracy (e.g., O’Kane Reference O’Kane1993: 170, 191; Vanhanen Reference Vanhanen1997: 40), and because, in a number of respects, it is parallel to the approach of Przeworski and collaborators discussed later.
Is Concept Formation Inherently Based on Classificatory Reasoning?
Central to Sartori’s view of concepts, as formulated in his classic article on “concept misformation” (Reference Sartori1970), is the argument that concept formation is inherently based on classification and that dichotomies are therefore fundamental to reasoning about concepts. We focus on two parts of Sartori’s argument. First, he suggests that the process of human reasoning that underlies concept formation involves thinking in terms of classification and cut-points. Thus, “human understanding – the way in which our mind works – requires cutoff points that basically correspond (in spite of all subsequent refinements) to the slices into which a natural or qualitative language happens to be divided” (Sartori Reference Sartori1970: 1038). Second, Sartori applies essentially the same argument to norms of scholarly inquiry. He states that “whatever their limits, classifications remain the requisite, if preliminary, condition for any scientific discourse” (Reference Sartori1970: 1040).
With regard to the first point, since the time of Sartori’s formulation of his position in 1970, a large body of research in linguistics, cognitive psychology, and cognitive science has yielded strong empirical evidence supporting a more multifaceted view of human cognition. Although classification is fundamental to human cognition, reasoning about gradations is likewise fundamental. Linguists employ terms such as cline, scale, and pragmatic scale to characterize graded understandings, and even prior to the full development of the contemporary field of cognitive linguistics, Lakoff’s (Reference Lakoff1973) work on “hedges” pointed to the complexity of gradations away from central instances of a concept that are established by the application of modifiers to nouns. Rosch (Reference Rosch, Rosch and Lloyd1978) provided an empirical demonstration of the centrality of prototypes (i.e., cases that are understood as exemplifying a concept) in conceptual reasoning, and of graded reasoning in relation to prototypes (see also Lakoff Reference Lakoff1987). It is further argued that, far from providing an unreliable and uncertain foundation for human cognition, a system of thought centrally organized around ideas of gradation is more stable, flexible, and reliable in the face of changing empirical reality (Taylor Reference Taylor1995: 53–54). Overall, we are convinced that viewing human understanding as fundamentally anchored in classification presents an incomplete picture that fails to capture the remarkable capacity of the mind to conceptualize different modes of gradation and different forms of the partial occurrence of any given phenomenon.
Sartori’s second point concerns norms of social science inquiry. Here we are likewise convinced that the reasoning about similarities and differences that underlies concept formation encompasses not only ideas about sharp contrasts and cut-points but also ideas about different forms of gradation and ordering. Both are fundamental to the conceptual work entailed in formal measurement theory. The axioms and conceptual reasoning that are a logical underpinning for measurement encompass simultaneously ideas about equal versus nonequal, which provide the foundation for categorization, and ideas about greater than versus less than, which provide the foundation for graded reasoning (Roberts Reference Roberts1976: 476–78; Michell Reference Michell1990: 166–70). Both are also fundamental to less formal procedures for reasoning about conceptualization and operationalization. The choice of an approach based on either classification or gradation in fact involves a complex process of simplifying the available information about the cases under consideration. Although the “output” of this choice is presented in terms of cut-points or gradations, the “input” routinely includes observations and intuitions both about sharply defined contrasts and about gradations.
This conclusion has an important implication for Sartori’s well-known argument that “concept formation stands prior to quantification” and that “the progress of quantification should lag – in whatever discipline – behind its qualitative and conceptual progress” (Reference Sartori1970: 1038). To the extent that Sartori is asserting that the process of assigning numbers to cases should be preceded by careful conceptual reasoning, he is correct. But it is important not to link this essential assertion to his other claim that classificatory reasoning based on cut-points is necessarily the first step in concept formation. We emphasize instead, following the argument made earlier, that such conceptual reasoning is specifically not limited to classificatory thinking.
Object Concepts and Bounded Wholes
Sartori (Reference Sartori1987: 182–85; Reference Sartori1993: 118–20) has subsequently developed a more flexible argument about dichotomies that focuses on “object concepts” and “bounded wholes.” He begins by distinguishing between “contraries” and “contradictories,” a distinction he uses in conjunction with an approach that explores the meaning of concepts by examining their opposites. Sartori states that in the case of conceptual oppositions that are contraries, intermediate positions exist. Examples are big versus small, hot versus cold, and rich versus poor. By contrast, in the case of oppositions that are contradictories, there is no intermediate possibility. The opposing concepts “are not only mutually exclusive but also exhaustively exclusive” (Reference Sartori1987: 182). Sartori’s formulation thus echoes the traditional idea in philosophical logic of the “excluded middle” (Honderich Reference Honderich1995: 256–57). He offers alive versus dead, married versus single, and biped versus quadruped as examples of contradictories, and he argues that democracy versus nondemocracy should be treated as a contradictory (Reference Sartori1987: 182–84). Sartori’s distinction represents a flexible approach in that it validates the use of graded comparisons in treating some conceptual oppositions, while underscoring the importance in other instances of the dichotomous treatment entailed in contradictories.
In discussing contraries and contradictories, Sartori (Reference Sartori1987: 182–85) utilizes a related distinction between “object concepts” and “property concepts.” This distinction, in effect, provides a rationale for treating certain concepts as contradictories and hence approaching them dichotomously. When a concept is construed as an object concept, then it designates what Sartori refers to as a type or an entity, and applying such a concept to empirical cases involves “identifying an entity” (Reference Sartori1987: 183). The type in question may be a complex phenomenon, such as a given form of political system. Sartori argues that such systems are constituted by multiple attributes, all of which must presumably be present for a case to be classified as an instance of the concept. Thus, it can be thought of as a bounded whole. When an object concept is applied to a particular case, one must establish, in dichotomous terms, whether or not the case corresponds to the concept. Sartori contrasts the idea of object concepts with a property concept approach, in which the concept is viewed as a characteristic that cases display to varying degrees and that hence calls for a graded treatment (Reference Sartori1987: 183–84; see also Reference Sartori, Sartori, Riggs and Teune1975: 28–29).
We wish to underscore two contributions of this approach. First, Sartori explicitly presents it as a flexible perspective that allows for alternative ways of approaching the logical treatment of concepts. Concept formation thus involves “deciding which logical treatment is appropriate for what purpose” (Reference Sartori1987: 185). Relatedly, he discusses how particular concepts are construed as being suitable for one or the other of these logical treatments (Reference Sartori1987: 183; see also Reference Sartori, Sartori, Riggs and Teune1975: 29). Choice is involved, and the choice could presumably be influenced by the goals of the investigator. We believe that this pragmatic emphasis on the goals of research takes the discussion in a good direction. Second, Sartori’s argument about object concepts is promising because it focuses attention on the idea of an interaction among the component attributes of democracy. In the third section (“Toward More Specific Justifications”), we address this interaction via the distinctively conceptual idea that each component attribute can potentially take on a different meaning, depending on the presence of other attributes.
At the same time, it is essential to sound a note of caution. We have a vivid label for the mistake of overstating the degree to which the attributes one seeks to conceptualize cohere as if they were like an object – this error is “reification.” We present two observations concerning the potential problem of reifying bounded wholes, one about what is involved in “naming” phenomena, the other about changing empirical knowledge and evolving scholarly usage.
The observation about naming concerns a tacit belief, or what cognitive scientists call a folk theory, that people routinely hold about concepts. It is widely believed that, in relation to many phenomena in the real world, it is possible to identify an inherently correct name or system of names. People believe it is possible to identify “the name of a thing, the one that tells what it ‘really is’” (Brown Reference Brown1958: 17; see also Lakoff Reference Lakoff1987: 9, 118–21). This tacit belief about naming has both an upside and a downside. On the upside, if a specific name is understood as designating what the phenomenon under discussion “really is,” the use of this name can facilitate communication and make it easier to remember the argument being presented. Indeed, if the analyst has in fact meaningfully summarized a complex body of information as consisting of two distinct types, this impressive analytic accomplishment deserves to be clearly communicated. On the downside, if a particular name resonates primarily owing to this tacit belief, rather than because it provides an analytically appropriate slicing of reality, then this name can become a slogan that is employed in a sloppy and uncritical manner, with serious risk of reification. One possible consequence could be that the idea of bounded wholes is uncritically embraced for the wrong reasons.
Our other basic observation about bounded wholes and reification is that in the face of changing social reality, shifting definitions of the subject matter, and evolving theoretical understanding and empirical knowledge, conceptualizations that initially serve to justify a dichotomy based on a particular cut-point can subsequently break down. This breakdown can be illustrated with some of Sartori’s examples of contradictories. Social change in recent decades has, in many different ways, surely made married versus single a more complex distinction. The scholar studying evolving social relationships and related political and legal debates must consider alternative cut-points in defining married versus single as a dichotomy. Evolution in medical technology has led to complex legal, ethical, and policy debates about the definition of alive versus dead, as well as to the introduction of intermediate concepts such as “brain dead.” It can be argued that for chimpanzees, the idea of both biped and quadruped is relevant. More broadly, in many domains of knowledge, what initially appeared to be clear distinctions sometimes break down and are reconceptualized. Zoology has seen a long history of distinctions that initially appeared clearcut but that subsequently were superseded (e.g., Gould Reference Gould1983). Overall, it can be argued that in a variety of disciplines, (1) dichotomies are of central importance, (2) the cut-points that establish dichotomies may evolve over time, and (3) scholars face an ongoing choice between retaining the use of dichotomies, based on these potentially evolving cut-points, or shifting to a graded approach that employs multiple cut-points.
Shifting the Burden of Demonstration
In light of these issues of changing social reality and evolving theoretical understanding, we suggest a pragmatic approach that places Sartori’s distinction between object and property concepts on firmer ground. We carry even further Sartori’s idea, noted earlier, of “deciding which logical treatment is appropriate for what purpose.” It may often be the case that Sartori’s distinctions provide a useful way to characterize a specific approach to the formulation and definition of a concept. Thus, a scholar may well develop a particular understanding and application that precludes intermediate alternatives, making it appropriate to view the concept as if it designated an object consisting of a set of interrelated parts, all of which are treated as definitionally necessary. On the other hand, the broad claim that a concept itself necessarily has object-like characteristics involving interrelated parts may be far more problematic.
Indeed, conceptual disputes are often only deepened when different groups of scholars involved in the dispute each treat their own particular approach as if it were a valid overall characterization of the concept. In her influential study of representation, Pitkin demonstrates how recognizing the multiple competing views of a concept, each of which mistakes its own “partial view” for the “complete structure,” can help us to understand and overcome conceptual confusion (Reference Pitkin1967: 10–11). A variety of other accounts of the complexity of concepts likewise supports this conclusion, including Gallie’s (Reference Gallie1956) discussion of the multifaceted structure of many important concepts, Lakoff’s (Reference Lakoff1987) theory of the cognitive models that provide complex sources of conceptual meaning, and Freeden’s (Reference Freeden1996: 60–67) view of conceptual “morphology,” in which the potential meaning of a concept encompasses a spectrum of “quasi-contingent” features. Hence, scholars should be cautious in claiming to have come up with a definitive interpretation of a concept’s meaning. It is more productive to establish an interpretation that is justified at least in part by its suitability to their immediate research goals and to the specific research tradition within which they are working.
This flexible, pragmatic approach shifts the burden of demonstration in justifying choices about concepts. Rather than treating the distinction between object concepts and property concepts as a permanent status, we understand this distinction as depending on the specific meanings and definitions that particular authors give to a concept, and on the goals and context of research.
Examples of Generic Justifications
We now consider justifications for gradations or dichotomies that have been advanced by prominent authors in the comparative literature on democracy and democratization. These authors focus most of their arguments on generic claims that the concept of democracy inherently requires one approach or the other. These examples suggest to us that this dispute is hard to resolve at the level of these generic claims, which is a further motivation for shifting the focus of justification to more specific arguments.
Justifying Gradations
Bollen and collaborators have strongly aligned themselves with the argument that the distinction between democracy and nondemocracy should be viewed in graded terms. Over a series of publications, Bollen (Reference Bollen1980, Reference Bollen1993; Bollen and Grandjean Reference Bollen and Grandjean1981; Bollen and Jackman Reference Bollen and Jackman1989; Bollen and Paxton Reference Bollen and Paxton2000) has developed graded measures that treat political democracy as a property that regimes display in varying degrees. He incorporates gradations into his definition, defining democracy as “the extent to which the political power of the elite is minimized and that of the nonelite is maximized” (Reference Bollen1980: 372; see also Reference Bollen1993: 1208).
Bollen defends his choice of a graded approach with the generic conceptual assertion that it is dictated by the concept of democracy. Bollen and Jackman describe the use of a dichotomy as “hard to justify” because the resulting measures fail to reflect the “inherently continuous nature of the concept of political democracy” (Reference Bollen and Jackman1989: 617, 612). Likewise, Bollen asserts that “the concept of political democracy is continuous” and hence we “unnecessarily compromise the concept by considering it a dichotomous phenomenon” (Reference Bollen1990: 13).
Bollen builds on this conceptual claim to make a further argument in favor of gradations based on a concern about measurement error (Reference Bollen1990: 14). He argues that, although both approaches introduce substantial measurement error because they fail to capture fully the continuous nature of political democracy, there are reasons to favor a graded approach. Bollen believes that “we can do better than dichotomous or three-point scales” and that a dichotomy is “crude” relative to the ordinal scales we can produce (Reference Bollen1990: 18). He favors scales with more than two or three categories because the “error introduced by analyzing ordinal indicators as if they were continuous generally is less, the greater the number of categories” (Reference Bollen1990: 18). His argument about error obviously depends on his conceptual argument, given that his assessment of the error entailed in ordinal versus dichotomous approaches focuses on how well they can capture a phenomenon he presumes to be continuous. Bollen specifically rules out statistical procedures that assume the underlying concept entails distinct categories. He considers them “not appropriate since the concept is continuous” (Reference Bollen1990: 18).
Dahl, in his influential writings on democracy, likewise incorporates gradations into his conceptualization and definition. He argues that “countries vary enormously in the extent to which their governments meet the criteria of the democratic process” (Reference Dahl1989: 233). In Polyarchy, he adopts a view of democracy that defines it in relation to an ideal type and thereby lays the foundation for his graded approach. Hence, as “one end of a scale, or a limiting state of affairs, it can (like a perfect vacuum) serve as a basis for estimating the degree to which various systems approach this theoretical limit” (Reference Dahl1971: 2).
Having used the term democracy to refer to this ideal, Dahl introduces the term polyarchy for discussing actual regimes. Following Dahl, Coppedge and Reinicke (Reference Coppedge and Reinicke1990) have employed Guttman scale analysis to operationalize a graded treatment of polyarchy, and Dahl’s preference for gradations is also reflected in his use of their scale (Reference Dahl1989: 241).
Dahl is quite clear about the inadequacy of a classificatory approach. His conceptualization of polyarchy focuses on contestation and inclusiveness. He states that “since a regime may be located, theoretically, anywhere in the space bounded by the two dimensions, it is at once obvious that our terminology for regimes is almost hopelessly inadequate, for it is a terminology invariably based upon classifying rather than ranking” (Reference Dahl1971: 6).
Obviously, Dahl and Bollen have both made outstanding contributions to the study of democracy. They offer carefully formulated definitions that incorporate a graded approach, and their use of the concept appropriately follows their own definitions. Yet establishing a definition based on gradations does not preclude the possibility that other scholars will employ conceptualizations and definitions based on a dichotomy, and it does not provide a basis for adjudicating the dispute over democracy and dichotomies.
Justifying a Dichotomy
Sartori argues that the distinction between democracy and nondemocracy should be treated as dichotomous (Reference Sartori1962: 150–52; Reference Sartori1987: 156, 182–85, 205–7; Reference Sartori1991: 248–49; Reference Sartori1993: 118–20). Hence, the essential initial task is to establish exhaustive and mutually exclusive categories of democracy and nondemocracy (Reference Sartori1987: 182–85). In Sartori’s view, approaches that treat the difference between democracy and nondemocracy as a matter of degree are fundamentally flawed. “What is completely missed by this degreeism, or continuism, is that political systems are systems, that is, bounded wholes characterized by constitutive mechanisms and principles that are either present (albeit imperfectly) or absent (albeit imperfectly)” (Reference Sartori1987: 184).
Sartori’s claim about bounded wholes might appear to state an empirical hypothesis according to how imperfectly these mechanisms and principles are present or absent. However, our reading of Sartori’s position is that he invokes empirical cases to illustrate a generic conceptual assertion that is treated as a necessarily valid claim about the concept of democracy, rather than to offer a validation of a potentially falsifiable claim. Sartori illustrates the need for dichotomies by pointing to sharp empirical differences – for example, between the US and the Soviet Union, and between regimes that are so harsh as to produce large numbers of political refugees and those that receive these refugees (Reference Sartori1987: 184, 185). In these examples, he does not focus on the problem of comparing intermediate cases, among which the idea of gradations of democracy might potentially be especially relevant. Thus, he does not appear to look for evidence that would call his bounded-whole thesis into question.
Despite his arguments against a gradation-based approach, Sartori does not preclude the use of gradations altogether but argues that such a treatment should be applied only to countries deemed democratic in terms of an initial dichotomy. Thus, “what makes democracy possible should not be mixed up with what makes democracy more democratic” (Sartori Reference Sartori1987: 156; emphasis in original). Sartori argues that both issues can be addressed in a single, integrated framework as long as the analyst follows a specific two-step procedure. First, regimes must be classified as democracies or nondemocracies. Then, only as a second step, a further set of criteria can be applied to those regimes deemed democratic by the initial dichotomy. Only with regard to these cases should we inquire as to how democratic they are (Reference Sartori1987: 182–83). Sartori asserts that “unless the two problems are treated in this order, the oxen may well wreck the cart rather than pull it” (Reference Sartori1987: 156).
Sartori does not propose his two-step procedure for all concepts but specifically for those conceptualized as bounded wholes. His assertion of the need for the two-step procedure hence presupposes his previous claim that democracy must be treated as a bounded whole. To be plausible, this two-step procedure would require a fuller elaboration of the bounded-whole idea and a defense of its appropriateness for a particular definition of democracy. In the third section (“Toward More Specific Justifications”), we suggest what such an elaboration and defense might look like.
Przeworski, Alvarez, Cheibub, and Limongi’s argument for their dichotomy is based both on a generic claim about how democracy should be conceptualized and on a more specific claim about measurement and the empirical distribution of cases (see the third section). Regarding the first claim, their definition of democracy requires the selection of the chief executive and the legislature through contested elections, the presence of more than one party, and the actual rotation of the incumbents out of office after a reasonable interval (Alvarez et al. Reference Alvarez, Cheibub, Limongi and Przeworski1996: 19). They argue that the graded approach is flawed because it fails to recognize that any regimes in which executive and legislative offices are not contested should not be considered democratic to any degree (Alvarez et al. Reference Alvarez, Cheibub, Limongi and Przeworski1996: 21; Przeworski et al. Reference Przeworski, Alvarez, Cheibub and Limongi1996: 52). To argue that there are borderline cases that call for a graded approach is, in their view, “ludicrous,” because in a carefully applied classification of regimes, the ambiguous status of a case can only reflect “bad rules” or “insufficient information” (Alvarez et al. Reference Alvarez, Cheibub, Limongi and Przeworski1996: 21–22). If a treatment of regimes fails to distinguish clearly democracies from nondemocracies, this does not undermine the attempt to apply a dichotomy; rather, it means that the scoring procedures need to be modified to remove the ambiguity.
Although Przeworski and collaborators thus make a broad claim about democracy, they go further than does Sartori in basing their justification on a specific understanding of the interaction among the attributes of democracy. Yet this justification still does not provide as complete an argument as we would like. For example, in our view it remains unclear why a regime that has competitive elections for the presidency, rotation in the presidential office, and more than one party – but lacks competitive elections for legislative office – is not at least partially democratic. This approach could also be helped by a specific justification of why each of the component attributes should be understood dichotomously, and not in terms of gradations.
Przeworski and collaborators do not rule out all graded comparisons among democracies but instead advocate the same two-step procedure as Sartori, with the first, dichotomous step essentially based on the idea of bounded wholes.
They argue that “while some regimes are more democratic than others, unless offices are contested, they should not be considered democratic. The analogy with the proverbial pregnancy is thus that while democracy can be more or less advanced, one cannot be half-democratic: there is a natural zero-point” (Alvarez et al. Reference Alvarez, Cheibub, Limongi and Przeworski1996: 21).
Przeworski and collaborators offer a criticism of Bollen and Jackman that explicitly challenges a basic assumption of gradation-based approaches. “Note that Bollen and Jackman (Reference Bollen and Jackman1989) are confused: it is one thing to argue that some democracies are more democratic than others and another to argue that democracy is a continuous feature over all regimes, that is, that one can distinguish the degree of ‘democracy’ for any pair of regimes” (Alvarez et al. Reference Alvarez, Cheibub, Limongi and Przeworski1996: 21; see also Przeworksi et al. Reference Przeworski, Alvarez, Cheibub and Limongi1996: 52).
For the purpose of comparing regimes that fall short of their standard, Przeworski and colleagues thus preclude the option of reasoning about them as partial democracies. Instead, they propose a separate conceptualization involving three dimensions of “dictatorship”: whether it is mobilizing or exclusionary, how many formal centers of power it has (executive, legislative, and parties), and whether it rules within a framework of law or in some more arbitrary manner (Alvarez et al. Reference Alvarez, Cheibub, Limongi and Przeworski1996: 16–19).
In principle, we are sympathetic to this two-step procedure. Nevertheless, it would be valuable to formulate a more sharply focused argument about why certain core attributes entailed in a particular definition must occur together if a regime is to be considered minimally democratic according to this definition. As with the arguments for gradations noted earlier, such an argument may support the choice made in a particular research project, yet it should not be interpreted as a generic prescription that applies across all conceptualizations of democracy.
Toward More Specific Justifications
We now turn to examples of justifications for dichotomies that are more specifically linked to the theoretical and analytic goals of the research and to the particular context being studied. We do not seek to cover all possible justifications but rather to illustrate some issues that arise in offering more specific justifications. An important counterpoint runs through the discussion. Although each of these justifications is initially meant to defend dichotomies, we find, in relation to each justification, a counter justification that favors the use of gradations.
A major goal of the literature considered here is to understand the causes and consequences of democracy. However, the goals most relevant to our discussion are more specific. We first consider the implications of two analytic concerns that have been central in much of the comparative literature on democracy, namely a focus on events and a focus on subtypes of democracy. We then examine justifications – based on the empirical distribution of cases and normative concerns – that we believe provide part of the underlying rationale for a dichotomous approach. Finally, we explore how the idea of bounded wholes could be more fully developed as a justification for a dichotomy, and we consider some implications of a concern with achieving sharper analytic differentiation.
Studying Events and Subtypes
Events. An important part of recent research on democratization is routinely called the “transitions literature” because of its concern with the event called a transition. The study of events has become an important focus in comparative social science (Abbott Reference Abbott1992; Griffin Reference Griffin1993; Sewell Reference Sewell and McDonald1996), and, as Riker (Reference Riker1957) demonstrated in his classic article on defining events, the analytically rigorous study of events requires establishing their boundaries in dichotomous terms. Correspondingly, O’Donnell and Schmitter (Reference O’Donnell and Schmitter1986: 6) define a transition as “the interval between one political regime and another” (see also Huntington Reference Huntington1991: 11). This definition in turn calls for a dichotomous approach that establishes the cut-point or threshold in relation to which the event of a transition to democracy is identified.
In studies based on complex comparisons across countries and over time, important problems can arise in establishing a threshold that meaningfully delimits the onset of democracy. Bollen and Jackman raise two concerns about such dichotomous treatments. First, they point to the difficulty of establishing conceptual equivalence among cases of democratization that occur in different historical contexts. This concern with equivalence centers on the fact that “the nature of political democracy (especially inclusiveness) has changed considerably over the past decades” (Reference Bollen and Jackman1989: 619; see also Markoff Reference Markoff1996: 4, 116–17). Second, for many cases, they question the feasibility of locating a point in time at which democracy, conceived dichotomously, began. They argue that “it is meaningless to claim that democracy was inaugurated in a given country on a single date,” and that “dating the inauguration of democracy conceived in binary terms is an inherently ambiguous task. In fact, it is an impossible task” (Reference Bollen and Jackman1989: 618, 619).
One plausible response to Bollen and Jackman’s first concern is to adopt a context-specific approach to conceptual equivalence. A central problem in establishing equivalence lies in the fact just noted, that the plausible agenda of “full” democratization has changed dramatically over time. What could be viewed as full democratization by the standards of an earlier period might be seen as incomplete democratization by later standards. For example, in the late twentieth century, universal suffrage and the protection of civil rights for the entire national population are routinely seen as essential features of democracy, whereas in the nineteenth century they were not (Huntington Reference Huntington1991: 7, 16). In light of this problem of equivalence, one solution is to compare regimes according to whether they have achieved full democratization in relation to the norms of the relevant time period (R. Collier Reference Collier1999: chap. 1; Russett Reference Russett1993: 15). This context-specific conceptualization allows for a dichotomous classification of cases distinguishing those that are fully democratized from those that are not.Footnote 4
The goal of conducting strong tests of hypotheses can provide a rationale for adopting this conceptualization, and hence for sticking with a dichotomy. Thus, the analyst may accept the standard of equivalence of full democratization according to the norms of the historical period, in part because it leads to the inclusion of what may be considered “inconvenient facts” (Weber Reference Weber, Gerth and Mills1958: 147) from the standpoint of a major hypothesis being entertained in the study. For example, Russett (Reference Russett1993: 15) wishes to include nineteenth-century cases in his tests of the democratic peace hypothesis because that inclusion pushes him to deal with a greater number of conflicts that could be interpreted as wars between democracies. Similarly, Ruth Berins Collier (Reference Collier1999: chap. 1) is skeptical about some arguments concerning the pivotal role of the working class in democratization, and to make her case, she wishes to include late nineteenth-and early twentieth-century cases that were the basis for classic arguments about the working class role.
A plausible response to Bollen and Jackman’s second concern, regarding the problem of establishing a single starting date for democracy, is to temper the idea of a strict dichotomy. Recognizing that important distinctions are lost by employing a single threshold (Paxton Reference Paxton1995, 4, 18–19), the analyst may focus on two or more successive thresholds, which potentially are crossed at widely separated dates. By conceptualizing democratization as a sequence of steps, rather than as a single event, this approach in effect introduces gradations. Examples are found in Rueschemeyer et al. (Reference Rueschemeyer, Stephens and Stephens1992: 160–62, 205, 304–08) and R. Collier (Reference Collier1999: chap. 1). This approach allows both the author and the reader to evaluate the implications of alternative thresholds for the findings of the analysis. Along these lines, Geddes’ comparative study of democratic transitions that have occurred since 1946 is another example of an analysis that reports the consequences of alternative thresholds (Reference Geddes1999: Table 1). We strongly endorse this practice.
Notwithstanding the importance of this concern about meaningful starting dates, we would reject a blanket pronouncement in favor of the universal use of gradations. Although Bollen and Jackman are correct to suggest that a focus on democratization as a single event is flawed for many cases, this does not imply that it is inapplicable in all cases. Those cases where democratization is an abrupt, rather than protracted, process might be adequately analyzed using a dichotomy that treats democratization as a single, well-bounded event.
Subtypes of Democracy. In some studies, the larger set of arguments advanced by the authors is concerned with assessing the causes or consequences of what may be called “classical” subtypes of democracy (Collier and Levitsky Reference Collier and Levitsky1997: 435). These subtypes are understood as corresponding to countries that are definitely democratic and that have some further differentiating attribute. The focus on these subtypes presupposes a concern with delimiting the set of democracies, within which the subtypes are differentiated.
A prominent example of this focus on subtypes is found in studies of the consequences for regime stability of parliamentary democracy, as opposed to presidential democracy (Stepan and Skach Reference Stepan and Skach1993; Linz and Valenzuela Reference Linz and Valenzuela1994; Sartori Reference Sartori1994). The inclusion of cases in this comparison strongly implies a dichotomous standard for establishing which countries are democratic and therefore can be considered instances, more specifically, of the parliamentary and presidential subtypes of democracy.
Another example is found in comparisons of what O’Donnell (Reference O’Donnell1994) has called “delegative democracies,” that is, regimes with strong presidencies in which the “horizontal accountability” of the executive to the legislature is attenuated. O’Donnell is convinced that among democracies, this delegative pattern has the consequence of eroding political institutionalization. He specifically defines delegative democracies as regimes that are above a basic threshold of democracy (Reference O’Donnell1994: 56). Hence, inclusion in the set of delegative democracy assumes inclusion in a larger, dichotomously defined set of democratic countries. Still another example is found in the democratic peace literature, which explores the effect of national regime types on the likelihood that countries will go to war with one another. Elman (Reference Elman1997) argues that political scientists need to specify the democratic peace hypothesis more carefully by looking at the consequences of particular types of democracy for international conflict behavior, and she likewise focuses on the parliamentary and presidential subtypes. Her analysis, like O’Donnell’s, is specifically concerned with understanding a set of countries that are democratic.Footnote 5
Research questions conceptualized in terms of democratic subtypes may appear to require a dichotomous approach, yet, as with events, an alternative is available. These studies could ask, for example, whether the consequences of a presidential versus parliamentary organization of legislative–executive relations vary according to the degree of democracy, with different patterns emerging in borderline cases as opposed to cases that fully meet some accepted standard for democracy. This is a distinct, but certainly related, question. Delegative democracy could be analyzed in the same way. This graded approach is in fact followed in Shugart’s study, which seeks to explain the emergence of parliamentarism versus presidentialism. At one point in the analysis he looks at the strength of one explanatory factor among both (1) countries that are either semidemocracies or democracies and (2) countries that are specifically democracies (Shugart Reference Shugart1999: Table 5). Shugart thus moves away from the underlying notion of democracy as a well-bounded type by conceptualizing it in graded terms.
In their writing on delegative democracy, democratic peace, and parliamentarism versus presidentialism, Stepan and Skach (Reference Stepan and Skach1993: 3), O’Donnell (Reference O’Donnell1994: 56), and Elman (Reference Elman1997) are quite specific in stating that they are concerned with regimes that are democracies. This focus clearly depends on a prior dichotomous understanding of democracy as a type. The question then becomes, how does one justify the choice between this dichotomous understanding and a graded alternative?
Underlying Justifications
We are convinced that these choices about the study of events and subtypes often rest on underlying assumptions regarding the empirical distribution of cases and normative judgments. By recognizing these assumptions and defending them, scholars could provide better justifications for these choices.
Empirical Distribution of Cases. One of the important ways in which the context of research affects choices about concepts involves the empirical distribution of the cases being studied. In a given context, scholars may observe a gap between democracy and nondemocracy, either across national units or over time within a given country, with observed cases generally either possessing or lacking most of the defining attributes of democracy. Given this gap, a dichotomy may provide an adequate summary of the empirical contrasts among cases. The use of a dichotomy in this context is not a conceptual assertion that rejects a graded approach as inherently flawed. On the contrary, the empirical hypothesis that regimes do cluster in this manner must be evaluated within a graded approach that can assess whether a gap exists. Thus, by looking for gradations, scholars may justify the conclusion that, for a given context, a dichotomy is good enough.
The idea of an empirical gap between democratic and nondemocratic regimes was of great importance to the recent literature on democratization, which was routinely concerned with relatively dramatic shifts in which many attributes of regimes changed in a relatively short span of time.
Correspondingly, this literature made extensive use of a dichotomous conception of democracy/nondemocracy and of the idea of a regime transition as a well-bounded event.
Arguments about the empirical distribution of cases have also been used to justify a dichotomous approach, based on a concern with measurement error. Przeworski and collaborators argue that a “more refined classification will have a smaller error” if the distribution of cases is unimodal and approximately symmetric, whereas “there is less measurement error when a dichotomous scale is used” if cases are uniformly distributed (Alvarez et al. Reference Alvarez, Cheibub, Limongi and Przeworski1996: 31). Furthermore, if the distribution is “u-shaped,” as they suggest the distribution of democracy versus nondemocracy tends to be, then the advantage of a dichotomy in terms of reducing error is even greater than it is for a uniform distribution (Alvarez et al. Reference Alvarez, Cheibub, Limongi and Przeworski1996: 31). Although Elkins (Reference Elkins1999) has raised questions about the way these authors treat error, this unquestionably is an argument in which the criterion of justification focuses on the empirical distribution of cases.
If the choice of a dichotomy can be justified on the basis of the empirical distribution of cases, then a change in this distribution can lead to a different choice. Huntington’s Third Wave treated democracy as dichotomous, though he noted a few intermediate cases (Reference Huntington1991: 12). Yet in a later article, he observes that “as formal democratic institutions are adopted by more and more diverse societies, democracy itself is becoming more differentiated.” He therefore sees the need to focus on a “democratic–nondemocratic continuum,” on which one finds “a growing number of countries somewhere in the middle” (Huntington Reference Huntington1996: 10). Diamond makes the same argument with reference to Latin America, where he finds that the shift toward democracy has made it “more fruitful to view democracy as a spectrum, with a range of variation in degree and form.” Owing to this changing empirical reality, treating democracy as something “merely present or absent” has become “a sterile perspective” (Diamond Reference Diamond and Farer1996: 53). More recently, on the basis of a worldwide comparison of regimes, Diamond (Reference Diamond1999: Table 2.4) shows that the proportion of intermediate cases has doubled between 1990 and 1997.
Although information about the empirical distribution of cases should play an important role in choices about dichotomies versus gradations, it should not be taken as the sole determinant of all such choices. It needs to be balanced against the potential value of sharp analytic distinctions, such as those offered by the classical subtypes discussed earlier, and against normative concerns that we discuss in the next section. In sum, we argue that although a graded approach is needed to adequately capture a highly uniform distribution, for more discontinuous distributions the choice between dichotomies and gradations remains open, and the various other considerations discussed throughout this chapter may play a decisive role.
Finally, in certain limited situations, a dichotomy may be justifiable regardless of the empirical distribution of cases. For example, in exploring certain research questions, the analyst’s treatment of democracy vis-à-vis nondemocracy may need to reflect the viewpoint of the individuals who are being studied. With such an actor-defined, emic approach to differentiating among regimes, a dichotomy would be justifiable independently of the empirical distribution of cases, provided that the actors whose behavior is being examined do in fact think of the world in terms of a democracy/nondemocracy dichotomy.Footnote 6 Starr’s recent discussion of the democratic peace hypothesis provides an example of such an argument. He interprets the causal mechanism at work in the hypothesis as including the idea that “each side must understand that the other is a democracy” (emphasis in original). Hence, Starr argues that analysts can best evaluate this hypothesis if they establish a “threshold point” above which the relevant political actors in other nations view a given country as democratic (Reference Starr1997: 129–30). In specific instances, such an approach may not always win universal agreement, yet in principle it is an appropriate justification.
Normative Evaluation. Normative concerns play an important role in comparative research on democracy (Dahl Reference Dahl1971: chap. 2; Sartori Reference Sartori1987: 7–8), and these concerns can provide another source of justification for a dichotomous or graded approach. Indeed, it appears likely that normative concerns lurk behind many arguments in favor of dichotomies, even though they are not made explicit. Although the idea of a fact–value distinction remains a familiar point of reference for many social scientists, we must recognize that the general choice of research topics, and more specific choices concerning what outcomes are explained and how they are conceptualized, routinely have a normative component. A study tends to be viewed as more important if it seeks to explain a humanly important outcome, and viewing an outcome as humanly important involves normative appraisal.
An example of a study that is careful and self-conscious in explicating the normative criteria that can underlie the choice of a dichotomy is the summary volume by O’Donnell and Schmitter (Reference O’Donnell and Schmitter1986) in the Transitions from Authoritarian Rule series. Their point of departure was explicitly normative (Reference O’Donnell and Schmitter1986: 5, 11, 13) and was formulated in light of the political and social parameters, as of the 1980s, faced by the Latin American and Southern European countries with which they were concerned. Given these parameters, the authors were convinced that a plausible target for advocates of regime transitions was a “procedural minimum” version of democracy that encompassed free and fair elections, universal suffrage, and broad protection of political and civil liberties. They labeled this constellation of features “political democracy” (Reference O’Donnell and Schmitter1986: 8, 13),Footnote 7 and they employed a dichotomous distinction to establish it as a democratic target. In establishing this target, they were deliberately positioning themselves in relation to scholars and political actors who aimed at a lower or higher standard.
Although normative concerns may lead scholars to adopt a dichotomous treatment and may serve to justify the choice of a particular cut-point, similar concerns can also motivate a graded approach. In Democracy and Its Critics, Dahl expresses concern that in the evaluation of regimes, a dichotomous approach may “impose upon the moral and empirical complexities of the world a false Manichean orderliness” (Reference Dahl1989: 316). A dichotomy is “empirically misleading” because it overlooks the fact that “countries below a reasonable threshold for full polyarchy are of extraordinary variety.” As a consequence, such an approach is “morally inadequate and likely to lead to inept policies” (Reference Dahl1989: 316). Dahl’s concern for flexibility and subtlety in normative appraisal is further reflected in his suggestion that such appraisal needs “to make judgments about the dynamics of change, and particularly the direction and rate of change,” because “even highly repressive regimes are not morally and empirically equivalent if their dynamics of change are radically different” (Reference Dahl1989: 316).
These two perspectives on normative evaluation reflect the different purpose of the authors. O’Donnell and Schmitter were writing in the midst of a dramatic, worldwide episode of democratization. They sought to identify what they saw as appropriate targets (neither too low nor too high) at which political actors should aim in pursuing democratization within this context. By contrast, Dahl is making a more general statement about normative evaluation that is not embedded in a specific historical episode, and hence his more flexible approach is also appropriate.
Further Issues of Justification
Developing the Idea of Bounded Wholes. Given the emphasis on bounded wholes in the conceptual discussions reviewed in the first two sections of this chapter, it is somewhat surprising that we have not found what we consider fully elaborated versions of this potential justification for a dichotomous, type-based conceptualization of democracy. We seek here to develop the notion of bounded wholes by focusing on conceptual interaction among the defining attributes of democracy.Footnote 8 Specifically, we explore the argument that if each component attribute within the definition is to be meaningfully understood as reflecting an aspect of democracy, then the other attributes must also be present. It seems likely that if such an argument is to be effective, it would not be a generic argument about democracy but rather would be tied to a specific definition based on a particular set of component attributes, and potentially also a specific context of analysis.
Although we have not encountered examples of full conceptual justifications along these lines, we have found partial examples in which authors do consider this kind of interaction among certain attributes of democracy. Specifically, they assert that the meaning of some defining attributes of democracy is changed if an additional attribute is not present. Thus, a “negative score” on one key attribute in effect cancels the meaning of a “democratic score” on others. Whereas the bounded whole criterion entails the expectation that all of the attributes are interrelated in this manner, this more limited thesis holds that the score on one of the attributes affects the meaning of the others. Thus, the arguments presented in this section are not in themselves full justifications for a dichotomy, but they are an example of the kind of conceptual reasoning that, if applied to all defining attributes, could provide such a justification.Footnote 9
One example is found in discussions of “electoralism” (Karl Reference Karl, Drake and Silva1986). This term is applied to regimes that hold elections in which substantial competition occurs, with uncertain electoral outcomes, yet where widespread violations of civil liberties continue to be a fundamental feature of political life. Various analysts argue that the electoral arena should not be seen as genuinely competitive and uncertain if civil liberties are not respected. Thus, one finds an interaction among these attributes of democracy, in that the absence of civil liberties specifically cancels the interpretation of the other attributes as being democratic.
Another example is found in discussions of the problem that some “democratically” elected governments lack effective power to govern.Footnote 10 In several Latin American countries, one legacy of authoritarian rule has been the persistence of “reserved domains” of military power over which elected governments have little or no authority (Valenzuela Reference Valenzuela, Mainwaring, O’Donnell and Valenzuela1992: 70). Hence, despite free or relatively free elections, civilian governments in these countries are seen by some analysts as lacking effective power to rule. In light of these authoritarian legacies, and often in response to claims that these countries are “democratic” because they have held free elections, some scholars sought to modify the definition of democracy by specifying as an explicit criterion that the elected government must have a reasonable degree of effective power. With this revised approach, some scholars have excluded countries such as El Salvador and Chile, during certain periods, from the set of cases classified as democracies, even though they held relatively free elections (Karl Reference Karl1990: 2; Valenzuela Reference Valenzuela, Mainwaring, O’Donnell and Valenzuela1992: 70; Loveman Reference Loveman1994: 108–13). This revised definition has received substantial acceptance (Huntington Reference Huntington1991: 10; Markoff Reference Markoff1996: 102–4), although there has not been full agreement on the treatment of specific cases (Rabkin Reference Rabkin1992: 165).
In the context of our discussion of the conceptual reasoning that could provide a full justification of a dichotomous approach, the point here is as follows: In cases where the elected government lacks effective power to rule, it is not valid to treat the other defining attributes of democracy (e.g., competitive elections) as meaningfully measuring the presence of democracy. The absence of effective power to rule does not merely make countries somewhat less democratic; it undermines the meaningfulness of the other defining attributes of democracy.
If the search for interactions among attributes is to be convincing, it is important that, in principle, the investigator be able to find negative cases, that is, instances in which such a conceptual interaction might be found but is not. An example appears in the debate over the observation that in many new democracies, elected presidents at times make extensive use of decree power, circumvent democratic institutions such as the legislature and political parties, and govern in a plebiscitarian manner that is seen as having strong authoritarian undercurrents. Such tendencies are addressed by definitions of democracy that include checks on executive power and hence exclude cases of weakly constrained presidentialism (Schmitter and Karl Reference Schmitter and Karl1991: 76, 87; Ball Reference Ball1994: 45–46).
However, this innovation has not been widely adopted. In this example, a crucial point is that these presidents are elected leaders. Hence, it might be appropriate to treat these regimes as meeting a minimal standard for democracy and to avoid any further adjustment in the definition – as long as they maintain presidential elections, a more or less viable legislature, and a general respect for civil liberties, and as long as opposition parties are not banned or dissolved. Scholars have considered the option of viewing the weakness of checks on executive power as invalidating the democratic characterization of these regimes, but instead they have concluded that it represented additional useful information about regimes that should be considered democratic (O’Donnell Reference O’Donnell1994: 56).
This discussion of checks on executive power is a useful negative example, in which scholars conclude that the absence of one attribute does not invalidate or diminish the meaning of other attributes in relation to the concept of democracy. Scholars concerned with refuting a bounded-whole approach could usefully devote more attention to making such arguments about the absence of this specific kind of interaction among attributes. Rather than endorsing or rejecting the bounded-whole approach in general terms, we urge scholars to carefully address themselves to the attributes of a particular definition of democracy, and to ask whether all the attributes display the kind of conceptual interaction that we have explored.
Sharper Differentiation. Achieving sharper, more fine-grained differentiation is an important goal in the comparative analysis of democracy, and a standard view of the advantages of a graded, as opposed to dichotomous, approach is that it more effectively promotes this goal. The practice of giving explicit names to categories that group together similar cases is also a means of pinpointing and differentiating crucial attributes of regimes. In this section, we explore two strategies that pursue the goal of sharper differentiation by combining gradations with named categories.
First, with an ordinal scale based on a limited number of categories, names can be given to the categories. An example is Dahl’s adaptation of the Coppedge-Reinicke scale of polyarchy. Dahl applies to the marginally democratic categories in this scale such names as “dominant party regimes” and “multiparty nondemocratic regimes” (Dahl Reference Dahl1989: 241). Diamond (Reference Diamond and Farer1996: 57) offers a similar ordinal scale, based on Freedom House data, in which the categories have names such as “partially illiberal democracy” and “semicompetitive authoritarian.” To the extent that these names meaningfully identify important empirical differences among the categories, this form of the scale may convey more information than does the scale without the names. For example, Dahl’s label “multiparty nondemocratic regime” conveys more information about the cases in this category than would a simple numerical score on his scale. He thus adds, in relation to the ordinal idea of “more or less,” sharper differentiation concerning “more of what.”
A second combined strategy that achieves sharper differentiation begins on the side of categories but incorporates the idea of gradations. As Collier and Levitsky (Reference Collier and Levitsky1997: 437–42) have shown, in the names of “diminished subtypes” such as semidemocracy, the adjective serves to cancel part of the meaning of democracy, creating a type that is less than fully democratic by whatever definition the author is using but that still retains some attributes of democracy. The subtype thus expresses the idea of a gradation away from democracy.
The use of diminished subtypes presents an interesting alternative to employing an ordinal scale. Consider, for example, three diminished subtypes formed in relation to standard procedural definitions of democracy, which routinely include universal suffrage, fully contested elections, and civil liberties. In relation to that definition, one finds diminished subtypes that characterize cases as missing one of these attributes. Thus, “male democracy” is used for cases that lack women’s suffrage; “controlled democracy” is used for cases with some important limitation on contestation, such as the banning of one political party; and “illiberal democracy” is used for cases where civil liberties are attenuated (Collier and Levitsky Reference Collier and Levitsky1997: 440). With each of these subtypes, one attribute is missing, but other attributes in this procedural definition remain present.
Consider how this same information would be conveyed using an additive scale of zero to three, based on the sum of a dichotomous (zero–one) version of these three component attributes. With the alternative configurations of attributes that correspond to these three diminished subtypes, the score in all three instances would be a two, conveying no information about which attribute is lacking. By contrast, if diminished subtypes are carefully employed, that information is clearly conveyed in the names of the subtypes. In this case, the concern with sharper differentiation parallels a concern, expressed by Gleditsch and Ward (Reference Gleditsch and Ward1997: 381) in their recent assessment of the Polity III data, that the use of aggregated scales can divert attention from important insights that emerge at a more disaggregated level. In sum, in relation to the idea of “more or less,” diminished subtypes convey sharper, more disaggregated differentiation regarding “less of what.”
However, along with this advantage, scholars should note a downside: This approach can encourage an undesirable proliferation of subtypes. If this occurs, the potential gains in sharper differentiation could be canceled by the conceptual confusion that may result.
Conclusion
The debate on democracy and dichotomies raises basic issues, faced by both qualitative and quantitative researchers, concerning appropriate standards for justifying choices about the formation and application of concepts. We have argued that justifications for the use of a dichotomous or graded approach are most productive when they focus on specific arguments about the goals and context of research. Throughout the discussion, a counterpoint emerged in which arguments that initially appeared to favor a dichotomy could, with modification, be compatible with, or even require, the use of gradations. This counterpoint reinforces our conviction that justifications should be as specific as possible, and that scholars should recognize that conceptual choices may prove more ambiguous than they initially appear.
We have shown that decisions about gradations versus dichotomies are often built into the framing of research questions. Research that focuses on democratization as a well-bounded event and on classical subtypes of democracy favors dichotomies. However, alternative ways of viewing events and subtypes are also available that allow for the introduction of graded notions. In relation to these conceptual choices, we hold that, although gradations are necessary in certain contexts, in other contexts the empirical distribution of cases or normative concerns may justify a dichotomy. Justifications based on the conceptualization of regimes as bounded wholes are also promising but have not been adequately developed. Finally, our discussion of the goal of sharper differentiation points to the value of combining gradations with named categories.
Our pragmatic approach, which recognizes that concepts, definitions, and operationalization may evolve with changes in the goals and context of research, should not be seen as neglecting an essential concern with standardization and rigor. We certainly do not favor an “epistemological anarchism” in which “anything goes” (Feyerabend Reference Feyerabend1973). Rather, the specific goals of standardization and rigor are most productively pursued in conjunction with a realistic focus on how other goals influence the use and application of concepts. For certain concepts, it is not plausible, and may even be counterproductive, to assume that the accumulation of knowledge requires that all scholars adopt a standardized meaning. Instead, for such concepts, it is more realistic to aim for an accumulation of knowledge grounded in mutual comprehension among scholars who self-consciously recognize their conceptual decisions as real choices. Thus, they are choices from a range of alternatives, which, although they are justified in light of certain context-specific criteria, still allow the scholar to recognize the validity of other decisions in other contexts.
Open access