The Economic Effects of Democracy and Dictatorship Chapter 2: Democracy

The Economic Effects of Democracy and Dictatorship Chapter 2: Democracy Carl Henrik Knutsen Department of Political Science, University of Oslo Contact: c.h.knutsen@stv.uio.no April 29, 2010 Abstract This chapter discusses the democracy concept. First, it presents previous literature, particularly investigating the differences between various institutional and substantive democracy definitions. The chapter then presents a novel suggestion for how to structure the democracy concept. A substantive definition, provided by Beetham (1999), is taken as point of departure. The discussion identifies seven second-level dimensions of the democracy concept and discusses the interrelations between second-level dimensions and how these relate to the (first-level) definition. Thereafter, the chapter briefly discusses how democracy should be measured, before it presents and evaluates some of the leading democracy indicators in the literature. 1

1 What democracy is 1.1 Democracy as a contested concept There is no consensus among political scientists on how to define democracy. Crick (2002, 1) calls democracy an essentially contested concept. First, there is disagreement over whether democracy is a dichotomous concept, or whether there are degrees of democracy (see e.g. Przeworski et al. 2000; Elkins 2000; Hadenius and Teorell 2005). Second, there is disagreement over whether democracy should be defined according to the existence of particular political institutions ( institutional definition ) or according to some underlying principles ( substantive definition ) (e.g. Schumpeter 1976; Przeworski et al. 2000; Beetham 1999; Knutsen 2010b). Third, even among those agreeing on either an institutional or substantive definition, there is disagreement over which particular elements should be included in, and excluded from, the concept. Should institutional arrangements that protect freedom of speech be included in institutional definitions? Should economic equality be included in substantive democracy concepts? Fourth, and this follows partly from the disagreements above, there is disagreement over the democracy concept s logical structure (Goertz 2005). Are there necessary or sufficient requirements for a regime to be considered democratic, or are there multiple, but related, factors that can ensure democracy without any one factor being necessary or sufficient? Different democracy definitions have different problems, either conceptual or regarding how easily they lend themselves to operationalization. Choice of definition, which is related to choices on the four dimensions identified above, is inherently plagued by trade-offs. A more stringent definition, likely easier to operationalize, may have less face validity and exclude relevant elements of democracy. In the presence of trade-offs, choice of definition is to a certain degree tied to subjective judgements. However, some definitions are more plausible than others. It is at least important for researchers to be aware of different definitions, their 2

benefits and drawbacks, how conceptual definition affects choice of operationalization and how choice of definition might affect empirical analysis. If democracy is a contested concept, so is dictatorship. The term dictatorship has been used quite differently by different researchers (and politicians), which generates confusion. One obvious solution is to provide a stringent definition of dictatorship. Here, dictatorship means the opposite of democracy. When dichotomizing regimes along a degree of democracy dimension, dictatorship is thus used as a category for all regimes not categorized as democratic; dictatorships are regimes that have a relatively low degree of democracy. Political regime types can be classified along several dimensions; degree of democracy is only one. The multiple dimensions indicates we should expect large institutional variation among both democracies and dictatorships. Therefore, several authors have separated between different types of democracies (see e.g. Lijphart 1999; Powell 2000; Shugart and Carey 1992) and different types of non-democracies, or as I call them, dictatorships (see e.g. Linz and Stepan 1996; Wintrobe 1998; Geddes 2003; Hadenius and Teorell 2007). Some dimensions have been explicitly identified according to specified criteria, which allows researchers to combine them with the democracy dimension and generate multidimensional classification schemes. When all n dimensions are dichotomized, this results in regime definitions with 2 n categories. Other classification schemes lack such logical stringency, but are results of inductively constructed ideal-typical regimes, based on empirically observed cases. This has resulted in several democracy with a prefix labels, such as illiberal- or quasi-democracy (Carothers 2002; Diamond 2002). Other classification schemes start out with clearly identified definitions and classification criteria, but which do not easily lend themselves to combination with the democracy dimension (see e.g. Tsebelis 2002; Bueno de Mesquita et al. 2003). Some of these can be seen as conceptual competitors to the democracy concept, rather than nuances of regime type classifications with democracy as one dimen- 3

sion. 1 1.2 Institutional definitions of democracy Let us consider in depth the democracy-dictatorship distinction. It is useful to start with considering whether democracy is defined institutionally or substantially. This choice has repercussions also for the other three fundamental questions raised above, regarding dichotomous versus continual definition, the boundaries of the democracy concept and the concept s logical structure. First, I briefly present various institutional definitions. I present the case for such definitions, before I present some counterarguments, and the case for a so-called substantive definition. I will not review the democracy concept s history. This is brilliantly done among others in Dahl (1989). 2 I focus on the most important alternative conceptions of democracy in modern political science, and restrict the discussion to representative democracy in large political units, more specifically the modern state. 3 Rousseau (2004) famously denounced all forms of indirect representation if the will of the people were to be translated into policy; this is not a common position today. The Marxist notion of real democracy focused to a larger degree on societal power-relations and economic structures than superstructural political institutions. Nevertheless, as Diamond (1999, 8) points out, today s use of democracy focuses more strictly on political matters, leaving out social and economic components from the definition of democracy itself. The below discussion takes this as point of departure. 1 Nevertheless, the empirical correlation between degree of democracy and number of veto players and size of winning coalition is high. 2 Interested readers could confront for example Plato (2003), Aristotle (1981) and Machiavelli (1997). Rousseau (2004) can also be read as background for discussions on democracy and the general will. 3 See Dahl (1989, 1998) for discussions of direct democracy in smaller political units 4

1.2.1 The common good and the general will: Do they exist? A review of discussions tied to concepts such as general will and common good helps one understand the attractiveness of institutional democracy definitions. Schumpeter (1976, 250) famously claimed that the eighteenth century philosophy of democracy may be couched in the following definition: the democratic method is that institutional arrangement for arriving at decisions which realizes the common good by making the people itself decide issues through the election of individuals who are to assemble in order to carry out its will. Schumpeter s critique of this classical definition, in short, was that defining democracy according to institutions ability to ensure implementation of a general will or common good is problematic. These concepts simply lack a meaningful reference. Schumpeter (1976, 251) argued that for different individuals and groups, the common good is bound to mean different things. Nine years after Schumpeter, Arrow (1951) struck another blow at the general will. Arrow showed the impossibility of aggregating fixed, well-defined individual preferences to a determinate, well-defined collective preference, when the number of individuals and issue-dimensions increase sufficiently. There has been a lot of elaboration on Arrow s theorem, on its necessary conditions and on in what situations preference aggregation is possible (see e.g. Sen 2002). A large literature has also considered how specific institutional rules determine the outcome of voting and other preference aggregation mechanisms, and how players with control over the agenda can manipulate rules to achieve their personal desired outcome(see literature overviews and analysis in e.g. Shepsle and Boncheck 1997; Mueller 2003; Austen-Smith and Banks 1999, 2005; Persson and Tabellini 2000). I will not go into this literature and its implications for the democracy concept here. Note, however, that Schumpeter s and Arrow s criticisms of the general will has left most modern democracy theorists vary of defining democracy according to the possibility of achieving such an outcome. Some have tried to side-step Arrow s problem by considering the transformation rather 5

than aggregation of preferences. Democracy has been considered a system that allows for deliberation, with the possibility of reaching a final agreement on the common good in different issue areas. 4. I am not convinced by these efforts. Certainly, preferences on complex policy issues are not fixed for all voters, at least not before a proper discussion on alternatives and implications have taken place. 5 Then again, institutions that secure debate and allow citizens to reach enlightened understanding of issues are important requirements for democracy, also according to those not adhering to deliberative-common-good definitions of democracy (see e.g. Dahl 1971, 1998). Some voters may even be swayed by moral arguments to alter their meta-preferences (Elster 1989), for example by becoming more prone to accept just solutions rather than solutions that maximize personal income, although one should not assume this happens to often. Politics is inescapably linked to conflicts over resources between various actors, and preferred policies are often based on individual self interest (which is often furthered through the associations of like-minded). Moreover, experimental research indicates that even conceptions of justice are colored, perhaps subconsciously, by personal interest (Babcock and Loewenstein 1997). Even if citizens may experience preference convergence after deliberation, full convergence among all citizens seems unlikely on any issue. Even partial convergence may be impossible on some issues. Therefore, Arrow s aggregation problem remains, and a definition of democracy as a system that allows for the common good to be implemented is highly suspect. Habermas ideal theory of requirements for a non-coercive discourse yields valuable insight into how democratic deliberation could (and perhaps should) look like (see e.g. Habermas 1996), and Rawl s theory of justice (Rawls 1999) yield convincing boundaries for just institutional structures and some very general principles for just policies. However, even these elaborate frameworks, presented in idealized settings, yield no concrete, specific 4 See for example (Gutmann and Thompson 1996), and see Shapiro (2003, 10-34) for a review of the literature and a very good discussion 5 See for example Fishkin (1995) on deliberative polls. 6

policy implications. It is therefore unlikely that discussions among real-world individuals would unequivocally converge on one particular, specific and implementable policy that contributes to a common good or a just outcome, without coercion. Preference aggregation is unavoidable. Let us now return to the discussion of institutional democracy definitions. 1.2.2 Minimalist definitions: Democracy as regimes holding contested elections Schumpeter s alternative to a classical democracy definition was defining democracy, or more precisely the democratic method, as the institutional arrangement for arriving at political decisions in which individuals acquire the power to decide by means of a competitive struggle for the people s vote (Schumpeter 1976, 269). Thereby, Schumpeter by-passed the issue of general will by referring directly to a particular institutional mechanism, competitive elections. With this, he escaped the fuzziness, indeterminacy and even emptiness (democracy is impossible) of classical democracy definitions. More recently,przeworski has defended a minimalist, institutionally based democracy definition (see e.g. Przeworski 1999). 6 Przeworski et al. (2000, 15), for example, define democracy, simply, as a political regime in which those who govern are selected through contested elections. The benefits of minimalist, institutional definitions are analytical stringency and precision, which ease subsequent problems of operationalizing democracy for empirical research (Przeworski et al. 2000). Thus, perhaps the main argument for defining democracy institutionally is the need to move beyond intuition towards stringent empirical measurement. The conceptual definition is considered appealing, therefore, partly because it is easy to operationalize. Przeworski et al. s (2000) definition of democracy is exemplary in terms of stringency and logical clarity. A regime is classified as democracy if it passes four operational rules; a regime s failure to comply with one or more of these rules will put it in the dictatorship category. The first two rules relate to whether or not relevant offices are filled through 6 See also Przeworski et al. (2000, 13-35). 7

elections, and the following two establish whether eventual elections were contested. The first and second rulers are simply The chief executive must be elected and [t]he legislature must be elected (Przeworski et al. 2000, 15). The third rule says that [t]here must be more than one party. This rule is extended to consider whether governments used elections to subsequently establish no-party or one-party rule, or permanent electoral domination (pp. 20-21). The fourth, and most controversial, rule is the alternation rule. According to this rule, which applies to regimes that have passed the first three, regimes must prove themselves democratic by allowing for leadership change through election. Thus, at least one alternation of government after election must be observed for a country to be classified as democratic (pp. 23-28). As Przeworski et al. (2000) argue, it is difficult to reveal the democratic intentions of governments before they are tested by losing an election. Better then, they propose, to be careful with assigning the democracy label to regimes that have not proved themselves through actions rather than words. This results in Botswana being put in the dictatorship category, as the Botswana Democratic Party had yet to lose an election. If Przeworski et al. s book had been written some years earlier, Japan would also have been classified as dictatorial, given the Liberal Democratic Party s dominance until 1993. As the authors recognize, the stringent rules eradicate much of the subjective element in classifying democracies, but come at the cost of more Type II errors : Genuinely democratic regimes are classified as dictatorial because their governments have yet to experience an election loss. 1.2.3 Adding attributes and institutional requirements: Democracy as a multidimensional concept and lists of specific institutions Different criticisms can be raised against Przeworski et al. s (2000) minimalist position. First, we have more information about elections contestability than information on whether elections exist, the number of parties, and whether there is alternation. We can in principle observe restrictions on freedom of speech, election fraud, and other means of political 8

manipulation. Protection of civil liberties are relevant for contestability, since without such protection the government can control political discussion and processes prior to the election, which again influences the opposition s chances. Regimes can not score high on contestability if civil liberties are lacking (e.g. Diamond 1999, 8). Blatant violations of the election result at election day, like ballot box stuffing and manipulation of election lists, are also detrimental to contestability. Why not utilize all relevant information before passing judgements on whether a regime holds truly contested elections? There is perhaps a trade-off between including all relevant information on the one hand and the degree of subjectivity in judgements on the other.przeworski et al. (2000), however, seem to have prioritized avoiding subjectivity too much. The lists of institutions that guarantee competitive elections typically are assumed to include guarantees of freedom of speech, freedom of association, independent electoral commissions, and perhaps also an independent judiciary that can settle eventual election disputes. Summed up, rather than only relying on observations of actual government changes for measuring contestability, one should look for predefined institutional guarantees, or at least helping conditions, that enhance competitiveness. The argument above also indicates elections are insufficient for securing democracy, even when contested elections is the key criterion. One needs additional institutional guarantees, and the democracy definition needs to be broadened. One paradigmatic list of such institutional guarantees comes from Robert Dahl s Polyarchy (Dahl 1971, 3). This list includes for example freedom of expression and alternative sources of information. More generally, some requirements, like freedom of speech seem to figure in most lists of institutional guarantees, but others, like laws on campaign financing and restrictions on money in politics are only included by some authors. Thus, there is no consensus on the exact make-up of the list of institutions guaranteeing elections competitiveness. This is probably one of the reasons why Przeworski et al. (2000) opt for the observational outcome of actual government change through elections when scoring regimes. 9

One may argue the equation of democracy and elections make for a fallacy of electoralism (Diamond 1999, 9). Many researchers view also for example civil liberties as minimum requirements for a democratic system of government. As seen above, specifying the set of requirements for what makes elections competitive also lead to the inclusion of some additional institutions and civil liberties as requirements for democracy. However, many scholars consider not only the competitiveness of elections, when stressing the significance of for example freedom of association or speech. They rather focus on a set of dimensions or attributes (see e.g. Munck and Verkuilen 2002) of democracy, where contestability of political elite selection is only one. With these extra dimensions comes extra institutional requirements for democracy. Some authors are relatively explicit about which institutional requirements are needed to ensure a high value on each dimension (see e.g. Dahl 1998), but others are not (see e.g. Goertz (2005); Munck and Verkuilen (2002) for reviews on this point). The most common additional democratic attribute is the degree of participation in political processes (or inclusiveness ) (see e.g. Dahl 1971). Competitive elections where only a fraction of the adult populace can participate produce competitive oligarchy rather than democracy (Dahl 1971). Democracy is thus argued to be at least a two-dimensional concept. Several measures of democracy draw on this two-dimensional structure (Vanhanen 2000; Coppedge and Reinicke 1991; Gasiorowski 1996). Universal suffrage and extensive citizenship rights are two institutional elements that arguably enhance scores on the participation dimension. Institutionally guaranteed voting rights for the entire, or almost entire, adult population is considered a hallmark of modern democracy. Thus, for example Great Britain was more democratic after each of its expansions of the franchise in the 19th and 20th century, than before (see Acemoglu and Robinson 2006). A system with competitive elections is for instance more democratic if both male and female franchise is ensured than if only males are allowed to vote. The list of democratic dimensions can arguably be expanded further. As discussed by 10

e.g. Munck and Verkuilen (2002), it matters also whether those elected are the ones actually exercising power over political agenda setting and decision making. If, like in present Iran, an independent entity dominates important aspects of political life and elected officials have little power, democracy is but a facade. Elected officials ability to exercise power over decision making is therefore argued to be an important aspect of democracy. This has been reflected in some democracy indicators (Bollen 1980; Arat 1991; Hadenius 1992; Marshall and Jaggers 2002). Several suggestions exist for which particular institutional structures may facilitate a high value on this dimension. One important suggestion is a politically controlled military, as strong, independent militaries have historically exercised substantial influence over politics, for example in Latin America (Smith 2005). Another suggestion is checks on the executive power, so that the executive does not dominate elected legislators (Marshall and Jaggers 2002). A third suggestion is the absence of a non-elected head of state with substantial political powers. The monarch was historically important in European countries, well after the invention of elected assemblies; the elected assemblies fight to reduce kings power over political matters was long and hard in many European countries (see e.g. North 1981; North and Weingast 1989). Some argue that a strong and independent judiciary with constitutional review rights may constitute an undemocratic feature, as elected legislators are restrained in their ability to legislate (see chapters in Przeworski and Maravall 2003). We will not discuss the relation between the judiciary and democracy here, but there is a vital pro-democratic role for a strong, independent judiciary in securing individuals political and civil rights against violations; the judiciary has a vital role in upholding democracy for example by securing liberties that are vital for the contestability of future elections (see e.g. Beetham 1994; Diamond and Morlino 2005). 7 The expansion of democratic attributes has led to concerns of inflating the democracy 7 See (Meredith 2007) for a nice historical description of how the Zimbabwean Supreme Court fought for Zimbabweans democratic rights. 11

concept. One solution is to separate between democracy defined according to a minimalist or electoral definition and the quality of democracy. Diamond and Morlino (2005, x- xxxi), for example, list different dimensions of democratic quality. They distinguish between five procedural dimensions of democratic quality and two substantive. The five procedural dimensions are 1) rule of law, 2) participation, 3) competition, 4) vertical accountability and 5) horizontal accountability. The two substantive dimensions are 6) civil and political freedoms and 7) political equality. The authors also include 8) responsiveness as a link between the procedural and substantive dimensions. As will become clear below, I am skeptical of such a solution. The reason is that a proper democracy definition implies that some of these qualitative dimensions are highly relevant for degree of democracy. When using a substantive definition, indeed, it can be argued that the separation of degrees of democracy and quality of democracy is non-existent (see e.g. Knutsen 2010b). 8 Leaving the quality-debate aside, the expansion of democracy-relevant attributes also means the institutional requirements necessary for democracy is expanded. The more attributes are added, the more evident it becomes that democracy is not only tied to what happens at election day, but to the entire event-chain from pre-election political discussions, via election day, to the post-election implementation of policies. Degree of democracy can be reduced at each step. Several different institutional requirements are thus needed for a regime to be a democracy, some relevant for the conduction of campaigns, and other for for example transparent and non-corrupt implementation of policies. Dahl has been one of the preeminent scholars when it comes to identifying institutional requirements for democracy (e.g. Dahl 1971, 1989, 1998). For example, Dahl (1998) provides six answers to the question 8 This argument is indeed strengthened by Diamond and Morlino s definition of a high-quality democracy, which is a democracy that provides its citizens a high degree of freedom, political equality, and popular control over public policies and policy makers through the legitimate and lawful functioning of stable institutions (Diamond and Morlino 2005, xiii). As will become clear below, these elements are either included in a substantive definition of democracy (political equality and popular control), or they are are potential outcomes affected by degree of democracy (legitimacy and stability). 12

What political institutions does large-scale democracy require? (pp. 85-86). Dahl lists 1) elected officials, 2) free, fair and frequent elections, 3) freedom of expression, 4) alternative sources of information, 5) associational autonomy and 6) inclusive citizenship. 1.3 Substantive definitions of democracy Lists of attributes, or alternatively lists of specific institutional structures, are sometimes presented as a multi-dimensional democracy definitions. However, these lists do not identify what democracy is, but rather identify crucial elements of what a functioning democracy requires. These lists are not proper as a core definition of democracy, but play a central role at the second- (attribute), and even third (indicator), concept levels (see Goertz 2005). This is recognized by Dahl (1998). Dahl argues that the six above-mentioned institutional categories are necessary requirements for large-scale democracies, because they help realize five core criteria for democracy, which are effective participation, equality in voting, citizens enlightened understanding of political matters, control over the political agenda and inclusion of adults in the political process (pp. 37-38). Further, Dahl answers the pressing follow-up question [W]hy these criteria?, with the answer each is necessary if the members... are to be politically equal in determining the policies of the association (Dahl 1998, 38). By digging deeper and deeper, we recognize that lists of institutions are not what democracy is, but rather they are requisites for realizing certain attributes, which again enhance degree of democracy. What is democracy? is best answered with a substantive definition, a definition that points to some core principles. Thereafter, we can ask what attributes are relevant for realizing these core principles. Then, we can ask which institutional and other requirements that are needed, individually or in conjunction, for realizing a high value on these attributes. As indicated above, there are problems with defining democracy institutionally. A substantive democracy definition takes the populace s role in political decision making as point of departure, rather than specific institutions. One proponent of substantive democracy 13

definitions is Beetham (e.g. Beetham 1994, 1999). According to Beetham, it is problematic to consider democracy merely as a matrix of various institutions, and even rights. The pressing question is why particular institutions and rights are considered democratic? How can we answer that question without invoking a circular argument? According to Beetham, [t]he only way to avoid circularity is by specifying the underlying principles which these institutions embody or help to realize, and in terms of which they can plausibly be characterized as democratic (Beetham 1999, 90). Furthermore, Beetham claims that [t]he core idea of democracy is that of popular rule or popular control over collective decision making (Beetham 1999, 90), and adds political equality as a second criterion. This is the definition of democracy used in this study. 9 Conversely, dictatorships are regimes with lacking popular control over collective decision making and large political inequalities. Notice that both political inequality and popular control over collective decision making are continuous concepts. Thus, there are also degrees of democracy, when defining it substantively. Moreover, as no observed societies have come even close to fulfilling these criteria perfectly, and no society is ever likely to do so, the defining characteristics of democracy are best viewed as ideals. This point was recognized by Dahl (1971) who used the term Polyarchy to distinguish empirical regimes with a sufficiently high score on the degree of democracy dimension(-s) from the idealized regime at the positive pole(-s). Here, an ideal democracy is considered a regime with completely equitable distribution of possibilities to exercise control over public decision making among all citizens. The negative pole, the idealized dictatorship, is a regime where one man controls all political decisions. As is true with an idealized democracy, such a regime has never existed. 10 I will not produce an elaborate positive defence of the choice of underlying principles 9 There are other interesting attempts at providing definitions of democracy that starts with plausible underlying principles. For example, Tilly (2007) defines democracy as XX. 10 See Bueno de Mesquita et al. (2003) and Besley and Kudamatsu (2007) for interesting analysis on the role of a dictator s supporters for policy making in dictatorships. 14

in Beetham s definition. 11 The interested reader should rather confront Beetham (1999). At one point, it becomes difficult defending underlying principles, as one is forced either to ground them in an even deeper set of principles which may lead to a regress, argue in circle, or claim the principles are intuitive, axiomatic starting points. This is the Münchhausen s trilemma, well known in philosophy. A defense can be made for these principle along the lines of the last suggestions, namely that popular control over politics and political equality seem intuitive, axiomatic criteria for whether a regime is a democracy or not. These criteria are often either implicitly or explicitly invoked in several studies. Anyhow, the alternative to a substantive definition seems to be a minimalist democracy definition, or to keep adding attributes and institutional requirements in an inductive, and often unsystematic fashion, to keep the democracy definition in accordance with ones intuitive understanding of what an empirical democracy should look like. Both alternatives have serious drawbacks. The proponents of substantive democracy definitions thus argue that institutions in themselves do not make or equate with democracy. Institutions only contribute to democracy if they enhance the underlying democratic principles. One can still be interested in particular institutions when studying democracy, but these are only instruments that underpin democracy. An important advantage of substantive democracy definitions is that they do not miss the political reality behind the formal and observable structures of government (Grugel 2002, 22). The existence of a particular institution is not the only relevant factor for degree of democracy; also how particular institutions actually function are important. There may 11 At first glance, this substantive definition might look relatively similar to the classical definition scorned by Schumpeter (1976). However, it is not. This definition does not define democracy according to any particular outcome, but looks at the populace s opportunities for controlling politics, and their status in terms of political resources (political equality). Thus, one can still accept Arrow s theoretical result, and in instances where plausible preference aggregation can be performed (for example on one-dimensional issues), investigate whether democracy affects the chances of achieving the Condorcet-winning policy alternative. By not referring to a general will the definition therefore evades some of Schumpeter s criticisms of classical democracy definitions. However, the vagueness and indeterminacy criticisms still apply to a certain degree: What does actually popular control mean, and who is the relevant demos? Addressing these criticisms must be done at the concept s second (attribute) and third (indicator) levels. 15

also be non-institutional factors affecting democracy. For example, acts of election violence or coercion of legislators by small groups would reduce degree of democracy, according to the definition above, as such acts reduce equal control over political decision. Africanist scholars have been particularly skeptical of focusing solely on formal political structures when studying politics. The African state has been characterized as a shadow state (Reno 1995), and given a host of other characteristics, indicating that real political decision making and resource allocation take place outside formal institutions, for example through informal patron-client networks (see e.g. Clapham 1996, 249 256). African politics is informalized, and the undermining of state institutions can be used as a tool by political elites to further their own power and interests (Chabal and Daloz 1999). This points to the importance of considering informal structures when classifying political regimes; formal institutions, like multi-party elections and constitutionally guaranteed freedom of speech, do not necessarily ensure a high degree of democracy. The reason why a substantive democracy definition is fruitful in this regard, is that the analyst can always ask: Does this political practice (we observe in country x) contribute to or reduce popular control over political decision making or political equality. Important criticisms of substantive definitions is that they are vague and difficult to operationalize. However, even if contested elections are easier to operationalize than popular control over public decision making under political equality, is it necessary to let the operational definition determine choice of conceptual definition? I argue below, that this is not the case. Rather than starting with a clear, logically well-structured operational definition and then choose a corresponding conceptual definition, my approach is to start with the conceptual definition, ask what attributes are related to this definition, and then ask how institutions and other political practices affect the attributes. One still needs to strive for logical clarity, however. A precise, logical concept structure is vital, for example, to discussions of validity and reliability of operationalization. I will below discuss how particu- 16

lar second-level dimensions, or attributes, contribute to degree of democracy, as defined by Beetham (1999), and thereafter how particular institutional structures relate to the different attributes. 1.4 Discussion: The best of both worlds? Substantive definition and logical clarity Social science concepts often have multi-level structures, and more specifically a three-level structure (Goertz 2005). The democracy concept arguably has a three-level structure, with a core conceptual definition of democracy at the first-, supporting dimensions or attributes at the second-, and different, concrete institutions and political factors at the third level. Above, I agreed with Beetham that democracy is realized to the extent that such [public] decision-making actually is subject to the control of all members of the collectivity considered as equals (Beetham, 1994:28). These two principles of popular control over public decision making and political equality are the first level of the concept; the deeper principles (Goertz 2005). A sufficiently high score on both these dimensions are required for a country to be democratic; the first level of the concept is therefore structured logically as a conjunctural and proposition. However, these principles say little about how democracy is realized. Which features do we need to ensure popular control over political decision making and political equality? Some plausible answers, indicated above, are political contestation and participation. Such dimensions make up the second level of the democracy concept. However, answers to the question of how democracy is realized, in the last instance, crucially include specific institutional structures, like competitive elections, guarantees of freedom of speech and other institutional structures. These institutions are thus integral in backing up the attributes that constitute the concept s second level (Goertz 2005); they are facilitating structures that 17

help realize popular control over politics and political equality. In this view, the proponents of a Schumpeterian democracy definition have got it backwards. Competitive elections are not the underlying constitutive element of democracy; they are a vital empirical requirement to help realize democracy. A substantive democracy definition does not, therefore, necessarily downplay the importance of elections for democracy; it is difficult to ensure popular control over politics, at least in entities larger than small city states (Dahl 1989), without them. They are however not the defining characteristic of democracy. Also for empirical research, much is gained from operating with a substantive democracy definition with core principles at the first-, supporting attributes at the second-, and specific institutional and other political factors at the third level. One reason is that there are several institutional requirements many analysts consider important for a political regime to be democratic, in addition to elections. Adolf Hitler also held elections under the Third Reich, but few would call this regime democratic. One reply may be that these elections were not competitive, but then we at least need to know what makes elections competitive. We thus need some principles to decide which extra requirements are needed for elections to be competitive rather than uncompetitive. Moreover, politics does not stop after election day is over. An elected government may concentrate power to itself, restrict debate on policy formation, restrict people s access to provide input to or even learn about political processes. Arguably, such a regime does not score very high on degree of democracy, and we need definitional principles that clearly indicate why. This means that also broad institutional democracy definitions, and not only minimalist Schumpeterian definitions, are problematic. The reason is that it is difficult to answer: What institutional characteristics should be included or excluded in the democracy definition?, if there are no overarching principles to judge their relevance by. Why is for example protections of freedom of speech regularly included in broad institutional definitions? It is preferable to be explicit about the steering principles that determine which empirical, institutional characteristics that are 18

found important or necessary for a democracy. Explicitly stating the underlying principles has some practical implications. First, the lists of democratic institutions vary widely between different authors. An a priori agreement on overarching principles enables a more thorough evaluation and discussion of the different lists relative merits. Second, political elites have proved to be very creative, perhaps more so than democracy researchers. The menu of election manipulation is long (Schedler 2002a,b), and researchers may be forced to add institutional requirements to their democracy definition, as history rolls along, in an ad hoc fashion. This is less problematic with a substantive definition, where the definition is history-independent, but where adjustments are systematically made at the third level of the concept, according to the overarching steering principles at the first level, and various attributes identified at the second. 12 Above, I identified the logical structure of the first level of the democracy concept as conjunctural. The logical structure of the second level is more complex, and will be discussed thoroughly below. The third level, as mentioned, relates to the practical operationalization of democracy in terms of measurable characteristics, such as existence of institutions. On this level, Goertz (2005) argues, the logical relation between the indicators is often not one of necessity, but of family resemblance combined with a logical or structure. There are several different types of electoral systems that can be designed to satisfy the second level principle of competitiveness, and there are several distinct ways to de facto guarantee alternative information to the public. When it comes to the practical design of institutions that enable democracy, there are thus multiple paths to the promised land. Consequently, operationalizing democracy properly is a difficult task, and as Munck and Verkuilen (2002) argue, there does not exist a fully satisfactory cross-country measure of democracy at present. 12 Notice that such a definition also allow us to compare degree of democracy over historical periods, such as ancient Athens versus present day Greece. It can easily be argued that the first is found wanting in terms of political equality and therefore is less democratic. We thus escape relativism and the argument that different epochs were associated with different institutions, and thus hard to compare. 19

If we opt for a substantive definition of democracy, the strong case put forth by Przeworski et al. (2000) for a dichotomous definition vanishes. Dichotomous measures have been attacked on several grounds, also because they may generate less reliable measurement (Elkins 2000). 13 However, most researchers react to the dichotomous definition because it seems implausible, and lacks face validity. Most democracy researchers probably have an intuition that Sweden is more democratic than Venezuela or Russia, but that these in turn are more democratic than North Korea. However, the logical defense of Przeworski et al. (2000) seems plausible, as they argue that contested elections is something you either have or not. Nevertheless, if democracy is defined as popular control over politics under political equality, it is easy to argue that democracy is continuous; there are different degrees of popular control and there are also degrees of inequality. Dahl (1971) was right that democracy is an ideal concept, a yardstick for empirical regimes to be compared against. No polyarchies are very close to the ideal, but some are closer than others. Whether we use Dahl s or Beetham s underlying principles, the continuous model of democracy with the endpoints being ideal types, is a good starting point for understanding what democracy is. 1.5 Sketching the democracy concept s structure Above, I discussed the underlying definition of democracy; the democracy concept s first level (Goertz 2005). As I discuss large, complex societies where political representation is necessary (Dahl 1989, 1998), popular control over politics relates to the question: To what extent are citizens able to elect their desired representative and, most importantly, ultimately affect the policies that are being decided upon by politicians? 14 Notice that logically, the populace can elect representatives without affecting policies and they can affect policies 13 The choice of dichotomous versus continuous measure is not innocuous either, as it impacts on empirical results regarding the causes and effects of democracy (see e.g. Hadenius and Teorell 2005). 14 See Parry and Moyser (1994) for a discussion on direct democracy, representation and substantive democracy concepts. 20

even if there is no election, although the interdependence is arguably strong. When it comes to political equality, this is not a completely separate dimension from popular control, as it relates to the distribution of control over collective decision making within the broader populace. For analytical purposes, we will however sometimes discuss popular control and political equality separately. Then, popular control refers to the ability of, or degree to which, broad segments of the population to influence actual decision makers (the politicians). Political equality, when discussed distinctly from popular control, relates more specifically to the distribution of power to elect and affect political decision makers within the general population (non-politicians). The two first level dimensions are not orthogonal, but can to a certain extent be meaningfully discussed separately when addressing their relations to the second level dimensions. We need to list the second level dimensions (attributes) that enhance democracy, and explain how they enhance democracy. Goertz (2005) elaborates on the importance of focusing on how different concept-levels relate to each other. Are for example the different second level dimensions necessary and jointly sufficient in bringing about democracy, as defined at the first level? This has implications for the logical-mathematical structure of the concept. Then we also need to elaborate on interrelations between the different second level dimensions. In their discussion on democratic quality dimensions (Diamond and Morlino 2005) are quite explicit about the systemic nature of democratic quality, and notice the strong interrelations of some dimensions. Many of these dimensions are here used as dimensions of democracy rather than democratic quality, although the distinction may be mostly semantic. Inglehart and Welzel (2006) provide the most explicitly modeled interrelation between second level dimensions in the democracy literature. They do so when discussing effective democracy, which is quite similar to the substantive democracy concept used here. They argue that formal political and civil rights are rendered less effective when there is weak rule of law (which is assumed to lead to a high degree of political corruption). Despite the prob- 21

lematic, oversimplified actual specification of relation between the dimensions (non-weighted multiplicative), and a host of measurement problems at the operational level (see Knutsen 2010b), the theoretical discussion is innovative and points to an important and overlooked aspect of democracy: The interrelatedness of different second-level dimensions. We will get back to what these interrelations may look like. First, however, let us first list the relevant second-level dimensions, and very briefly argue why they are relevant. 15 1.5.1 Seven dimensions It is important to explicitly recognize the multi-dimensional nature of political regimes (see e.g. Gates et al. 2006). However, there are differences among theorists on what dimensions to include in and exclude from a democracy concept. I argue that the following dimensions should be included: Competition (C), Participation and Political Inclusiveness (P), Political and Civil Rights (R), Horizontal Accountability (H), Vertical Accountability (V), Rule of Law (L) and Political Effectiveness (E). Consider democracy, D, as a function of these seven dimensions; that is D = D(C, P, R, H, V, L, E). Elaborating on the nature of this function is a nice, precise way to discuss the concept s structure. All the first-order derivatives are arguably positive, that is D X > 0, where X refers to any of the seven dimensions. In words, degree of democracy increases when there is a positive change on any of the seven dimensions. 16. Let me explain briefly why these first-order derivatives are positive. As discussed above, Competition (C) is the key element in minimalist democracy def- 15 On this point, the literature is quite extensive: (see e.g Dahl 1971, 1998; Beetham 1994, 1999; Diamond 1999; Diamond and Morlino 2005; Inglehart and Welzel 2006; Schmitter and Karl 1991; Munck and Verkuilen 2002; Goertz 2005; Tilly 2007) 16 It can be discussed whether this is the case at very high initial levels on the R- and E dimensions. Various chapters in Przeworski and Maravall (2003) discuss the relationship between democracy and rule of law, and we noted above some problematic (democratic) aspects related to a strong judiciary and constitutional review rights. Moreover, assuming that D L > 0 may actually be quite problematic when values on other dimensions are very low. That is, we have an empirical dictatorship which enforces its laws strictly. Whether consistent enforcement of laws generated by a dictatorial elite leads to more democracy is questionable, to say the least. Nevertheless, Inglehart and Welzel (2006) have made a strong case for the argument that improved rule of law increases effective democracy when there is a minimum level of political and civil rights, but that there is no effect when there are no formal rights 22

initions, and is a vital ingredient in any democracy concept-specification. Competition refers to multiple entities contesting for the ability to formulate and implement various sets of policies. The system that arguably has proved best able to secure competition in large societies, is multi-party elections. Elections allow citizens to choose between alternative visions of how to organize society and alternative policy-menus. Elections also have other important functions. Importantly, they function as disciplining devices for non-myopic politicians, when these consider taking actions where their and the electorate s interests diverge (see e.g. Ferejohn 1986). The existence of elections may more generally alter the policy platforms of parties towards positions that are favored by a larger part of the citizenry (Downs 1957) These two latter points will be treated separately under the vertical accountability dimension below. Here we focus purely on the direct effects of elections, related to the citizenry selecting the representatives they want to legislate and implement policies. Competition is directly related to popular control over collective decision making. If we accept that representation is practically necessary in large-scale societies, choosing between alternatives, for example parties or candidates, is perhaps citizens most important device for controlling policies. Everything else equal, the introduction of one extra candidate weakly increases the value of the most preferred policy platform for any citizen. The institutionalized choice between multiple contending political elites thus separates democracy from dictatorship. For competition to function smoothly, elections should be held relatively frequently with regular intervals. Competition is not ensured by an non-contended elected government occupying office for decades. Citizens can change their minds, old citizens die and new citizens should gain the right to choose between policy alternatives. Plus, regular competition has a disciplining effect on politicians. The nature and quality of implemented policies hinge not only upon proposed legislation (during campaigns), but the ability and willingness of politicians to work hard, pass legislation, and ultimately ensure implementation. These factors depend not so much upon the 23