Chapter 1 from book manuscript Democracy on Retreat: Crisis, State-Society Relations, and the Recurrence of Autocracy in the Interwar Years Agnes Cornell Assistant Professor, PhD Department of Political Science, Aarhus University Bartholins Allé 7 8000 Aarhus C, Denmark E-mail: [email protected] Jørgen Møller Professor, PhD Department of Political Science, Aarhus University Bartholins Allé 7 8000 Aarhus C, Denmark E-mail: [email protected] Svend-Erik Skaaning Professor, PhD Department of Political Science, Aarhus University Bartholins Allé 7 8000 Aarhus C, Denmark E-mail: [email protected] 1 Chapter 1: Patterns of Democratic Breakdown Revisited: A Crisp Perspective Introduction Since the advent of modern democracy there has only been one large-scale democratic rollback if we measure this as an absolute fall in the tally of democracies (Møller & Skaaning 2013: Chapter 5). This rollback took place in a hugely interesting context, namely the interwar period. It is therefore unsurprising that the issue of democratic breakdown and survival in the interwar period has been subjected to a number of comparative studies (see Berg-Schlosser & Mitchell 2003; Bermeo 2003; Capoccia 2005; Gusy 2008; Karvonen 1993; Luebbert 1991; Overy 1994; Rueschemeyer et al. 1992). However, these comparative studies have been hamstrung by the poor quality of cross-national data on regime change. Possible the best illustration of this is the fact that scholars disagree vehemently about which countries were democratic during the interwar period – and, ipso facto, about the number of instances of democratic breakdown and democratic survival. This can be illustrated by comparing the extant attempts to score interwar spells of democracy based on categorical differences between democratic survival and breakdown. Presently, four general democracy datasets do this (see Bernhard et al. 2001; Boix et al. 2013; Marshall et al. 2013; Vanhanen 2003).1 On top of these general datasets, In Table 1 and 2 we include two more 1 Bernhard et al.’s (2001) scoring is premised on Dahl’s (1971) two dimensions of democracy (polyarchy), i.e., competitiveness and inclusiveness. It therefore includes four criteria: that the political regime is characterized by competitive elections without vote fraud affecting the general outcome; the absence of extensive or extreme violence that inhibits voters from expressing themselves; that there is no banning of political parties representing a substantial portion of the electorate; that suffrage is extended to at least fifty percent of adult citizens. Boix et al. (2013) identifies years where the executive is directly or indirectly elected in popular elections and is directly responsible to voters or to an elected legislature; where the legislature (or the executive if elected directly) is chosen in free and fair elections; and where a majority of adult men has the right to vote. The Polity IV dataset (Marshall et al. 2013) builds on indicators on executive recruitment, political competition, and executive constraints that reflect the presence of institutions and procedures through which citizens can express effective preferences about alternative policies and leaders, on the one hand, and the existence of institutionalized constraints on the exercise of power by the executive, on the other hand. It follows from these requirements that Polity IV disregards the extent of suffrage. On this basis, a 21-point polity scale, ranging from -10 to 10, is constructed. We use the standard procedure, suggested by the people behind the Polity 2 qualitative overviews of democratic breakdown and survival in Europe and Latin America, respectively, constructed by Bermeo (2003: 23) and Mainwaring et al.’s (2008).2 We follow Lipset (1959) in sorting our cases into two categories: Europe and British settler colonies on the one hand and Latin American countries on the other hand. [Table 1 about here] With respect to the former category, we find quite some agreement across the different datasets. As illustrated in Table 1, virtually all the “old” democracies of Northwestern Europe and the British settler colonies produce little in the way of incongruities. Only France and Switzerland stick out in the data provided by Bernhard et al. and this is solely because they did not introduce universal (female) suffrage until after World War II. Moreover, a number of “new” democracies, such as Czechoslovakia, Estonia, Poland, and Ireland, are scored in consistent ways across the five accounts that cover Europe.3 However, in a large minority of European cases, there is no consensus about whether democratic breakdowns occurred at all. The most conspicuous disagreements concern the scoring of measure, to distinguish between democracies and non-democracies, i.e., a polity scoring of at least 6 on the polity scale is considered democratic. Vanhanen’s (2003) measure is a fine-grained scale, constructed by multiplying the ratio of opposition seats in parliament with the level of electoral participation. We use Vanhanen’s own suggestion for a threshold to make a distinction between democracies and autocracies. To qualify as democratic, a case needs to score at least 30 percent on the Competition indicator, at least 10 percent on the Participation indicator, and at least 5 on the Index of Democracy. 2 Bermeo, who largely bases her overview on Berg-Schlosser and Mitchell (2003), does not link her coding to an explicit definition of democracy or particular coding criteria so it is difficult to establish whether disagreements with her overview are due to differences in the definitions of democracy. But her overview is almost identical to those provided in other comparative overviews of democratic breakdowns in the interwar years, such as those of Capoccia (2005), Guzy (2008), Lee (1994), and Overy (1994). Regarding Mainwaring et al. (2008), we list the country-years identified by them as either semi-democratic (competitive) or democratic. 3 The small differences we find are mainly due to the fact that some datasets score countries based on the status at the beginning of a year whereas others choose the ending of a year to categorize. The year of a democratic transition or a democratic breakdown therefore sometimes differs. The first point can be exemplified by the Polish case where some dataset code 1918 as the first democratic year whereas others op for 1919. Similarly, in the Czechoslovak case some datasets identify 1938 as the year of breakdown with others opting for 1939. 3 six countries: Bulgaria, Italy, Lithuania, Portugal, Romania, and Yugoslavia. Bermeo considers all of these six countries as democratic at some point in the interwar period. The other accounts only assign a democratic status to some of these cases and, for several of the cases, the years tallied as democratic differ. On top of this, Spain and Finland also stand out because the Polity IV scores indicate that they experienced democratic breakdowns in 1923 and 1930, respectively, while the other accounts only identify a democratic collapse in Spain at the time of the civil war (1936-39) and score the Finnish case as a democratic survivor throughout the period. [Table 2 about here] Table 2 shows that the disagreement is even bigger with respect to Latin American countries. According to Bernhard et al., only Uruguay 1934-39 can be considered democratic in the interwar period. At the other end of the spectrum, we find Mainwaring et al.’s account, according to which there were no less than 19 democratic spells in 14 Latin American countries. The three other datasets can be situated along this spectrum – but with a tendency to align more with Bernhard et al. than with Mainwaring et al. Polity IV identifies only two democratic spells in two Latin American countries whereas Boix et al. recognize five spells in four countries, and Vanhanen eight spells in seven countries. If we disregard the very restricted scoring of Bernhard et al. and Polity IV, we can note that the Southern Cone countries of Argentina, Chile, and Uruguay are those most consistently highlighted as democratic, at least for part of the interwar period. On top of this, there is general consensus that Costa Rica was democratic for the entire interwar period. Next, there is some agreement that Columbia, Nicaragua, and Honduras had democratic spells. But other than that, the disagreements are rampant. 4 The need to set the record straight The inconsistencies in Table 1 and 2 are somewhat surprising considering that most of the included measures are based on a crisp distinction between democratic breakdown and survival. One would imagine that the relatively modest ambitions of these datasets make for high validity and reliability. To be sure, some of the incongruities are due to the fact that the distinctions are based on different conceptions of democracy, particularly with respect to suffrage criteria. Most importantly, Bernhard et al.’s conceptualization of democracy is more demanding than that of the other datasets on this point which is reflected in a more restricted scoring. That said, different suffrage criteria are obviously not the sole reason for the disagreements as indicated by the fact that Polity IV is one of the most restrictive datasets even though it does not take the extension of suffrage into account. Moreover, Bernhard et al. consider Bulgaria 1919-1920 as democratic, which the other indices do not. It is therefore clear that some of these cases must simply have been miscoded in some of the datasets. The suspicion about miscoding is supported by the relatively long democratic spells that we find in almost all columns of Table 1 and 2. Apart from the Italian democratic breakdown in 1922, there is only one instance of a short post-World War I democratic spell in Europe, namely Bernhard et al.’s aforementioned scoring of Bulgaria 1919-1920 as democratic. This runs counter to an established point in the historiography of especially East-Central Europe between the wars, namely that a virtually ubiquitous, genuine democratic competition in the immediate aftermath of World War I was repeatedly suffocated when popularly elected governments began to threaten elite interests (see Palmer 1970; Rothschild 1974).