Necessary and Sufficient Conditions for the Breakdown/Survival of Electoral Regimes and Fascist Takeover/Non-Takeover in the Interwar World

Transcription

1 1 Necessary and Sufficient Conditions for the Breakdown/Survival of Electoral Regimes and Fascist Takeover/n-Takeover in the Interwar World Steffen Kailitz Hannah-Arendt-Institute for Research on Totalitarianism at the University of Dresden Draft paper prepared for presentation at the Santiago 2009 World Congress of Political Science, Santiago de Chile, July Introduction Around 1918 political observers (e.g. Bryce 1921) thought they were watching a worldwide victory of democracy pretty much like at the beginning of the 1990s (e.g. Fukuyama 1992): It was assumed that with the spread of popular elections the principles and methods of democracy would eventually prevail among all enlighted people (Deat et al. 1934: 1). At both points in time a big wave of democratization ran through Europe. But soon during the interwar period an authoritarian reverse wave started. With the takeover of Lenin in Russia in 1917 the era of Communism and only a few years later in Italy in 1923 the era of fascism began. Until 1939 all the states of Central and Eastern Europe, with the exceptions of Czechoslovakia, Finland, and Switzerland had fallen to various forms of autocratic rule. The reasons for the survival and breakdown of electoral regimes in the interwar years appears to be, especially after the Interwar project of Dirk Berg-Schlosser and Jeremy Mitchell (2000, 2003, Berg-Schlosser/de Meur 1994, Berg-Schlosser 1998), the most studied issue in the subfield of democratization studies that focuses on the persistence of democracies (see Aarebrot/Berglund 1995, Ertman 1998, Linz/Stepan 1978, Skaaning 2008). So why does it make sense to analyse the causes of the survival and breakdown of electoral regimes in the interwar years again? First, different from Berg-Schlosser (2000, 2003) and other previous studies on the breakdown of electoral regimes in the interwar years I consider all countries in the world, not only the European countries. 1 I thank Erik Fritzsche, Tanja Havlin, and Katrin Strehle for research assistance.

2 2 Second, instead of restricting myself to a top-down analysis of the sufficient conditions of democratic breakdown and survival I started with a bottom-up-analysis of necessary and sufficient conditions and in addition to that analysed the sufficient conditions also top-down. All other previous studies which used csqca, fsqca and mvqca to examine the conditions of democratic breakdown and survival in interwar year did provide truth table solutions which were the result of a top-down analysis of the sufficient conditions, but many papers did not even examine the necessary conditions separately and none included the necessary conditions systematically in the solution formula. Third, electoral autocracies were excluded from the analysis. Fourth, even though the main focus of the paper is the breakdown and survival of electoral regimes in addition to that I analysed the conditions for a fascist takeover and non-takeover, too. The aim of this paper is to identify conditions that are necessary and/or sufficient for 1. the breakdown/survival of an electoral regime and 2. a fascist takeover/nontakeover in the interwar period by using Charles Ragins measures for sufficiency and necessity (Ragin 2006, 2008) and a bottom-up configurational analysis 2. A condition is completely sufficient for the outcome, when always if the condition is present the outcome takes place. The cases with the condition are a subset of the cases with the outcome. This does, however, not mean that the outcome cannot take place, when the condition is not present, because there can be other conditions or combinations of conditions leading to the same outcome. CONDITION IS COMPLETELY SUFFICIENT Condition absent Condition present Outcome present 1. not relevant 2. cases here Outcome absent 3. not relevant 4. no cases here A necessary condition is a condition that always is present, when the outcome is present. The cases with the outcome are a subset of the cases with the condition. But there can be cases where the condition is present but not the outcome. CONDITION IS COMPLETELY NECESSARY Condition absent Condition present 2 Configurational analysis is used as a generic term for qualitative comparative analysis, fuzzy set/qualitative comparative analysis and multi-value qualitative comparative analysis. See Ragin 2008; Ragin/Rihoux 2008.

3 3 Outcome present 1. no cases here 2. cases here Outcome absent 3. not relevant 4. not relevant 2. Selection of cases My first step is to define the universe of cases. To measure regimes a twodimensional scale seems necessary. On the first dimension the degree of democracy is measured. The two main dimensions of democracy are public contestation and participation (Dahl 1971). Tatu Vahanen (1990, 1997, 2003) constructed two measures competition and participation which correspond to those characteristics of political systems that best distinguish more democratic systems from less democratic ones. Competition is calculated by subtracting the percentage of the votes won by the largest party from 100 percent. If parliamentary and presidential elections are taken into account, the arithmetic mean of the two percentages is used to represent the degree of competition. Participation is calculated by the percentage of the population that actually voted in the elections. The number is calculated from the total population, not from the adult or enfranchised population. On this dimension the regimes were coded as follows: Electoral democracy (1): Participation > and Competition >= 30 Electoral oligarchy and/or semi-competitive system (2): participation < 15 and >= 5 and Competition < 30 and >= 10 n-participative and/or non-competitive regime (3): participation < 5 and competition < 10 Too much of the previous literature places an overwhelming emphasis on the electoral process and thus overlooks other fundamental dimensions that are criteria for analysing regimes (Snyder 2006: 220). But there are regimes with elections which are still autocracies (Levitsky/Way 2002, Schedler 2006). According to my definition an autocracy is a regime in which the executive is (almost) unrestricted and the citizens are not free. Mexico until the end of the 1980s is a classical example of an electoral autocracy. Only electoral non-autocracies are able to experience something like a democratic breakdown, so the autocracies have to be sorted out. On the second dimension I therefore measure the inverse degree of autocracy. The inverse measure of autocracy has been constructed by using the measure of

4 4 executive constraints by Polity IV (Polity IV Project 2009) and a measure of political freedom (free/partly free/non free) inspired by freedom house. The degree of freedom is evaluated on the basis of qualitative research (e.g. Linz/Stepan 1978, Berg- Schlosser/Mitchell 2000). Thus, I distinguish between three regime types on this axis: Liberal-constitutional regime (1): there are effective executive constraints and the citizens enjoy a great deal of civil liberties (in the words of Freedom House they are free ). Hybrid regime (2): there are at least some executive constraints and the citizens enjoy at least some civil liberties (in the words of Freedom House they are partly free ). Autocracy (3): there are almost no executive constraints and/or the citizens enjoy only few or no civil liberties (in the words of Freedom House they are not free ) In this paper I consider all liberal-constitutional and hybrid electoral regimes in the period from 1919 to 1938 with a population of more than inhabitants all over the world. So Liechtenstein, Luxembourg, Monaco and San Marino are expelled because of their small size. I further exclude all regimes which are either completely non-democratic or completely not liberal-constitutional. Hence, I exclude Romania and Hungary which were considered by Berg-Schlosser/Mitchell (2000, 2003) as well as the Soviet Union, Bulgaria and Albania which were considered by Skaaning (2008). Why did I exclude these countries? In Hungary, for example, in 1926 a white regime took over, led by István Bethlen, a Transylvanian aristocrat, and Miklós Horthy, the former commander in chief of the Austro-Hungarian Navy. It was a reaction to the red revolution, led by, Béla Kun. So, the red terror was followed by a white terror. Many Communists and other leftists were tortured and executed without trial. From 1932 on the Hungarian regime had showed strong fascist elements, but anyway at no point in the interwar years Hungary had been nonautocratic. There is only one dubious case. Honduras is classified as an electoral autocracy although the Polity IV data measures a hybrid regime in the interwar years because there had been only presidential elections and in addition to that there was massive electoral fraud (Bendel 1995: 150). Two data sets are constructed for the analysis. In the first set the electoral oligarchies are included, in the second set excluded. According to my definition electoral oligarchies independent of their score on the autocracy-dimension are regimes with a participation of less than 15 per cent of the population in the national elections. Among the non-autocratic electoral regimes in the interwar years Argentina, Cuba,

5 5 Portugal, Uruguay, and Yugoslavia have been electoral oligarchies. In this paper 16 electoral regimes Australia, Belgium, Canada, Costa Rica, Czechoslovakia, Denmark, Finland, France, Ireland, Netherlands, New Zealand, rway, Sweden, Switzerland, United Kingdom, and the USA which survived all the interwar years and 14 electoral regimes that broke down are considered. Only two countries Germany and Italy experienced a fascist takeover, in two other countries Austria and Spain fascist were included in the government. Table 1: Breakdowns of Electoral Regimes in the Interwar Years Country Date of Before change Change of regime breakdown of regime Italy 29-Oct-1922 Hybrid electoral to electoral democracy autocracy (3/1); on 03-Jan-1925 from electoral autocracy to non-electoral autocracy (3/3) Poland 12-May- Liberalconstitutionaelectoral to hybrid non system (2/3) Portugal 28-May democracy Hybrid electoral oligarchy to non-electoral autocracy (3/3) Fascist takeover Yes Fascist inclusion Yes Lithuania 17-Dec-1926 Hybrid electoral democracy Cuba 01-Jan-1928 Hybrid electoral oligarchy Yugoslavia 06-Jan-1929 Hybrid electoral oligarchy Argentina 06-v-1930 Hybrid electoral oligarchy Germany 30-Jan-1933 Hybrid electoral democracy Uruguay 31-Mar-1933 Hybrid electoral oligarchy Austria 25-Jul-1934 Hybrid electoral democracy Estonia 12-Mar-1934 Liberalconstitutional democracy Latvia 15-May- Liberalconstitutional 1934 democracy Greece 04-Aug-1936 Liberalconstitutional democracy Spain 01-Apr-1939 Liberalconstitutional democracy to non-electoral autocracy (3/3) to non-electoral autocracy (3/3) to non-electoral autocracy (3/3) to non-electoral autocracy (3/3) to non-electoral autocracy (3/3) to non-electoral hybrid regime (2/3) to electoral autocracy (3/1) to non-electoral autocracy (3/3) to non-electoral autocracy (3/3) to non-electoral autocracy (3/3) to non-electoral autocracy (3/3) Yes Yes Yes Yes 3. Method: The Analysis of Necessary and Sufficient Reasons

6 6 In everyday live we all know that necessity is at least as important as sufficiency, but the classical statistical approaches failed to identity necessary conditions. Only recently, researchers (Goertz 2006, Goertz/Braumoeller 2000, Goertz/Starr 2002) have shown an increased interest in the importance of necessary conditions. Boolean algebra is the optimal tool to check for the necessity of conditions (Ragin 2006, 2008). However, when Charles Ragin (1987) first introduced csqca he focused mainly on the case orientation of the method and its ability to analyse causal complexity. Even if it was told from the beginning that Boolean algebra is a tool to test the necessity and sufficiency of conditions (Ragin 1987: ) the main advantage of the method that was emphasized by Ragin and his followers was that it is able to discover different ways to the same outcome. So most researchers who use a configurational analysis until today be it csqca, fsqca or mvqca work top-down and search for sufficient conditions of the outcome. The researcher checks some possible conditions with fs/qca or Tosmana and the programme returns a minimized so called truth table solution. But this solution which one gets from csqca, fsqca and mvqca is only a list of different combinations of causal factors that have met specified criteria of sufficiency for the outcome to occur. The top-down approach is just able to identify sufficient conditions, but necessary conditions can only be checked bottom-up (Ragin 2000: 218). Consequently if the researcher looks for necessary and sufficient conditions he always has to start checking his conditions bottom-up one by one (see e.g. Schneider/Wagemann 2007: 57-62). 3 In addition to the identification of necessary conditions the way bottom-up has also some more advantages. It is more case-orientated than the way top-down. Condition by condition the researcher deals with the truth tables and it is easy to address which cases are behind a result. Another crucial advantage of a bottom-up approach is that one usually does not have to care much about the problem of limited diversity, which is a basic problem of a top-down configurational analysis (on this problem see Ragin/Sonnett 2005). In this paper the sufficiency and necessity of the different explanations for democratic breakdown/survival are first tested bottom-up one by one. After that I provide a topdown analysis of the sufficient paths to the outcome (the classical configurational 3 A problem when someone only uses the classical top-down-analysis of sufficient conditions via Tosmana or fs/qca is also that the algorithm sometimes identifies a condition as necessary, because it is included in all paths to the outcome. But sometimes this condition does not pass a bottom-up test of necessary conditions.

7 7 analysis). At last I will merge the solution formulas for necessary and sufficient conditions. To analyze the necessity and sufficiency of a condition I use Ragins measures of consistency and coverage of necessary and sufficient conditions. The measure of consistency tells us to which degree a subset relation has been approximated. The coverage assesses the relevance of a consistent subset (Ragin 2006: 2008: 44-70). With the measures of consistency and coverage it is possible to identify if a condition is necessary and/or sufficient for an outcome. The measures are calculated by the following formulas: consistency necessity (and at the same time coverage sufficiency) = number of cases with condition and outcome/number of cases with outcome consistency sufficiency (and at the same time coverage necessity) = number of cases with condition and outcome/number of cases with condition The consistency is always the crucial measure for sufficiency and necessity. It is pointless to look at the coverage if a condition is not close to be sufficient or necessary. For the measure of consistency a minimum requirement has to be chosen, while this is not necessary for coverage. Ragin (2008) suggests that the level of consistency should be above 0.8. I set a threshold of 0.9 to be on the safe side. But there is one point to emphasize: If it is spoken of necessity and sufficiency in the context of a configurational analysis it is only meant a possible necessity or a possible sufficiency. The fewer cases one analyzes the easier he can go wrong with his claims. When it comes to necessary conditions we always have to check if a condition is only trivial. Trival are all necessary conditions in which the cause is present in all cases, irrespective of outcome. Neither peace nor war can be without air and gravity. If there is no variation of the condition dependent on the outcome the necessity of the condition is only trivial (Caramani 2008: 62, Goertz/Braumoeller 2000: 854). We can also speak of trivial necessity if there is no variation in the outcome. What is the surplus value by measuring consistency and coverage of sufficient and necessary conditions compared to the classical quantitative toolkit? A correlation matrix reveals if there is a relationship between certain conditions and the outcome. Only if we have the very rare luck, to get a perfect relationship, the condition is

8 8 necessary and sufficient. But all values in between one and zero do not tell us if a condition is either (close to be) sufficient or (close to be) necessary. 4. Theoretical Explanations: Operationalization and Results 4.1 Structural Approaches Socioeconomic conditions Since Seymour Martin Lipsets seminal article Economic Development and Democracy (1960) the following thesis is one of the best known in social sciences: the more well-to-do a nation, the greater the chances that it will sustain democracy (31). The argument in a nutshell is that wealth could create widespread higher education, strengthen the middle class and create a democratic civil society and thus make democracy sustainable. Modernization is for Lipset the precondition of the development of a democratic political culture. The four factors which Lipset originally tested were wealth, urbanization, education and industrialization. I focus on wealth, namely GNP, and education, namely literacy. It can be argued that poorer democracies are in greater danger to break down as well as that democracies need educated people to survive. 4 Nevertheless, it does not make much sense to state that urbanization or industrialization foster democracy directly. For the purpose of this paper I construct an index of the degree of modernization of a country as follows: Highly developed country: High GDP (> 3750$) and High Literacy (> 90%) Medium developed country: Medium GDP ( $) and High Literacy (> 90%) or High GDP (> 3750$) and Low Literacy (< 90%) Low developed country: Low or Medium GDP (< 3749$) and Low Literacy (< 90%) All highly developed electoral regimes survived the interwar years (Australia, Belgium, Canada, Denmark, France, Netherlands, New Zealand, Switzerland, United Kingdom, and USA). Hence, a high development is a sufficient condition for the survival of an electoral regime with at least some basic constraints on the executive and at least some political liberties. The result stays the same if we consider the 4 For a more thorough argument why literacy is important for the democratization process see Hadenius 1992.

9 9 electoral oligarchies or not. But high development is not a necessary condition for the survival of an electoral regime since it survived in Costa Rica, Czechoslovakia, Finland, Ireland, rway, and Sweden without the condition. At the same time all electoral regimes that broke down were only medium or low developed countries (Austria, Estonia, Germany, Italy, Latvia, Lithuania, Poland, Spain, Greece, Portugal, Argentina, Cuba, Uruguay, and Yugoslavia). So the absence of a high development is a necessary condition for the breakdown of an electoral regime. In other words it means that an electoral regime will not break down, when it passes a certain level of development. But if I would only slightly reduce the threshold for a high GDP, Austria (GDP 1913: 3465$ per capita) and Germany (GDP 3523$ per capita) would also operate under the label of highly developed countries. That is why one could argue with Lipset himself as well that at least the German development contradicts the modernization theory and Germany is an example of a nation where growing industrialization, urbanization, wealth and education favored the establishment of a democratic system, but in which a series of adverse historical events prevented democracy from securing legitimacy and thus weakened its ability to withstand crisis (Lipset 1960: 28). However, it is obvious that in Germany, Austria and in (the electoral oligarchy) Argentina electoral regimes broke down despite being at least medium developed countries. On the other hand in Costa Rica the electoral regime survived the interwar years even despite a low degree of modernization. Still we should not forget that the electoral regime in Costa Rica broke down in 1917 as well as in 1947, so this case is not really a strong contradiction to the claim that electoral regimes with a low level of development broke down in the Interwar years. 5 In the case of Ireland that has a medium GDP we do not know about the literacy rate, so this case could be coded as well as having a medium development. Table 2: Modernity (high) and Breakdown Modernity Outcome QCA- Countries (high) Outcome 0 0 C Australia, Belgium, Canada, Denmark, France, Netherlands, New Zealand, Switzerland, United Kingdom, USA 0 1 Austria, Estonia, Germany, Italy, Latvia, Lithuania, Poland, Spain, Greece, Portugal, Argentina, Cuba, Uruguay, Yugoslavia Costa Rica, Czechoslovakia, Finland, Ireland, rway, 5 However in the present we can observe the pattern of a surviving democracy despite a low level of modernization for example in India.

10 Sweden Data management and analysis in this paper was performed using fs/qca 2.5 and Tosmana. Source for GDP in both tables: Maddison All the GDP data refer to Source for literacy: Vanhanen Coding: 1: High development; 0: Otherwise. Table 3: Modernity (low) and Breakdown Modernity Outcome QCA- Countries (low) Outcome 0 0 C Australia, Belgium, Canada, Czechoslovakia, Denmark, Finland, France, Netherlands, New Zealand, rway, Sweden, Switzerland, United Kingdom, USA 0 1 Austria, Germany, Argentina 1 0 C Costa Rica, Ireland 1 1 Estonia, Italy, Latvia, Lithuania, Poland, Spain, Greece, Portugal, Cuba, Uruguay, Yugoslavia Coding: 1: Low development; 0: Otherwise. The explanatory value of a modernization index which combines GDP and literacy is higher than the explanatory value for literacy or GDP alone. I do not agree that socioeconomic indicators only have a rather limited explanatory power (Berg- Schlosser/De Meur 1994: 257, see also Skaaning 2008). But still the level of development can not explain why the democracies in Czechoslovakia, Finland, Ireland, rway, and Sweden survived, while the democracies in Austria and Germany broke down Historical Structures Barrington Moore claimed in his seminal work Social Origins of Dictatorship and Democracy (1966) that there are three routes to modernity, a democratic (followed e.g. by England, France, and the USA), an authoritarian-reactionary and later fascist and a communist. Moore argued that the undemocratic paths are linked to the occurrence of a strong labour-repressive agrarian elite. In the footsteps of Moore Rueschemeyer/Stephens/Stephens (1992) and Stephens (1989) pointed out that a strong labour-repressive agrarian elite undermines democratic stability. By using data of scholars as Rueschemeyer/Stephens/Stephens (1992: ) and Erssons (1995) I distinguish between the presence (1) and absence (0) of a significant landed upper-class engaged in labour repressive agriculture. In some countries like the United Kingdom, Canada and New Zealand we find strong agrarian elites, but the agriculture is not labour-repressive. These cases are coded as 0.

11 11 Table 4: Strong Agrarian Elites and Breakdown strong Outco QCA- Countries agrarian elites me Outcome 0 0 Belgium, Canada, Denmark, France, Netherlands, New Zealand, rway, Sweden, Switzerland, United Kingdom C Australia, Costa Rica, Czechoslovakia, Finland, Ireland, USA 1 1 Austria, Germany, Italy, Estonia, Latvia, Lithuania, Poland, Spain, Greece, Portugal, Argentina, Cuba, Uruguay, Yugoslavia Sources: Ersson 1995, Rueschemeyer/Stephens/Stephens 1992, Stephens 1988 The absence of a strong agrarian elite combined with a labour-repressive agriculture is a sufficient condition for the survival of an electoral regime in the interwar years. And the existence of a strong agrarian elite combined with a labour-repressive agriculture is a necessary condition for the breakdown of an electoral regime. Again the result is the same if we consider the electoral oligarchies or not. That means an electoral regime survived the interwar years when there was no strong agrarian elite that repressed the landworkers and there was no electoral regime that broke down without a strong agrarian elite combined with a labour-repressive agriculture. If we take into account that some observers would accentuate that Germany had a significant landed upper-class engaged in labour repressive agriculture in the interwar years only in the Eastern part, the condition would fall a little short of sufficiency and necessity (for a different coding of the German case see Skaaning 2008). It is also obvious that the absence of a strong agrarian labour-repressive elite is not a necessary condition for the survival of an electoral regime (see the cases of Australia, Costa Rica, Czechoslovakia, Finland, Ireland and USA). That means electoral regimes can survive very well with a strong agrarian labour-repressive elite. In another seminal work on the influence of historical structures Stein Rokkan drew a conceptual map of Europe (Rokkan 1970, 1975, Flora/Kuhnle/Urwin 1999). Two axis divide the countries on his map. The East-West axis is based on the strength of city networks and political centre formation, that is strong in the East and weak in the West. The rth-south axis is based on the integration of state and church which was in Europe traditionally strong in the secularized Protestant rth and weak in the Catholic South (Aarebrot/Berglund 1995). In accordance with Aarebrot/Berglund (1995: 218) and Skaaning (2008) I distinguish between countries with an early state building and those which had not much of such experience prior World War I. This distinction is not related to the date when a country first appeared on the world map.

12 12 For example, Czechoslovakia and Finland were late nations in that respect. To Aarebrot/Berglund (1995) all Western and Central European countries had been exposed to state-building, on the contrary all the countries which derived from the Russian and Ottoman Empires had only very limited experiences in the process of state building. While the first group had experiences e.g. with feudalism and Roman law (Charlemagne heritage), the second group lacked these characteristics (Byzantine heritage). In addition to Aarebrot/Berglund (1995) I coded all the British and Spanish settler colonies like their motherlands Great Britain and Spain. As we can observe it is almost a necessary condition that a country is a secularized Protestant country that the democracy survived in the interwar years and it is almost a sufficient condition that a country was Catholic and non-secularized to experience the breakdown of an electoral regime. Nevertheless, the electoral regimes in Costa Rica and Ireland survived despite of their Catholic culture. And after World War II more and more democracies in Catholic countries were successful like in Italy. Moreover, electoral regimes broke down in Estonia, Germany, and Latvia with a secularized Protestant majority of the population. Table 5: Protestant/Secularized and Breakdown protestant/ Outcome QCA- Countries secularized Outcome 0 0 C Costa Rica, Ireland 0 1 Austria, Italy, Lithuania, Poland, Spain, Greece, Portugal, Argentina, Cuba, Uruguay, Yugoslavia 1 0 C Australia, Belgium, Canada, Czechoslovakia, Denmark, Finland, France, Netherlands, New Zealand, rway, Sweden, Switzerland, United Kingdom, USA 1 1 Estonia, Germany, Latvia Source: Barrett/Kurian/Johnson Coding: 1: Protestant secularized majority of the population; 0: Otherwise. A Charlemagne heritage is a necessary condition for a democratic survival. electoral regime survived the interwar years without it. The absence of a Charlemagne heritage is also a sufficient condition for the breakdown of an electoral regime. t a single electoral regime with a Byzantine heritage survived the interwar years. But Estonia, Latvia, Lithuania, Poland, Greece, and the former states of Yugoslavia are all democracies today and there is no reason to believe that they are breaking down any time soon. So the sufficiency of the Byzantine heritage for the breakdown of an electoral regime faded away over time. A Byzantine heritage was also not a necessary condition for a breakdown in the interwar years. Because in eight countries the electoral regime broke down despite a

13 13 Charlemagne heritage it was not a sufficient condition for democratic survival, too. Only if the electoral oligarchies are excluded from our cases a Charlemagne heritage is relatively close to be not only a necessary but also a sufficient condition for a democratic survival in the interwar years. Table 6: Charlemagne Heritage and Breakdown Charlemagne Outcome QCA- Countries heritage Outcome Estonia, Latvia, Lithuania, Poland, Greece, Yugoslavia 1 0 C Australia, Belgium, Canada, Costa Rica, Czechoslovakia, Denmark, Finland, France, Ireland, Netherlands, New Zealand, rway, Sweden, Switzerland, United Kingdom, USA 1 1 Austria, Germany, Italy, Spain, Portugal, Argentina, Cuba, Uruguay Source: Aarebrot/Berglund 1995, Skaaning 2008 and own coding. According to Gregory Luebbert (1991) in all societies where liberals were hegemonic an interclass coalition followed. Hence, for Luebbert pre-war liberal hegemony is a sufficient reason for the survival of democracy in the interwar years (8). two conditions are tested in this paper: First, I test the effect of an electoral tradition before War War I. Second, I check if a liberal-constitutional tradition before World War War I is a sufficient condition for a democratic survival. The absence of a liberal-constitutional tradition before World War I is a necessary condition for the breakdown of an electoral regime with a relative high coverage. At the same time a liberal-constitutional tradition is a sufficient condition for democratic survival also with a high coverage. So Luebbert has been right. On the other hand the absence of an electoral tradition before World War I is only a necessary condition for a democratic breakdown and a sufficient condition for the survival of an electoral regime if we exclude the electoral oligarchies (Portugal, Argentina, Cuba, Uruguay, Yugoslavia) from our analysis and the coverage is even if we exclude them lower in comparison to a liberal-constitutional tradition. So according to the data a liberalconstitutional tradition is more important to explain the fate of electoral regimes than an electoral tradition. Table 7: Democratic Tradition pre World War I and Breakdown Electoral Outcome QCA- Countries Tradition Outcome 0 0 C Czechoslovakia, Finland, Ireland, Sweden, Switzerland 0 1 Austria, Estonia, Germany, Italy, Latvia, Lithuania, Poland,

14 14 Spain, Greece, Portugal, Yugoslavia 1 0 C Australia, Belgium, Canada, Costa Rica, Denmark, France, Netherlands, New Zealand, rway, United Kingdom, USA 1 1 Argentina, Cuba, Uruguay Source: Vanhanen Coding: 1: Electoral Tradition before World War I; 0: Otherwise. Table 8: Liberal-constitutional Tradition pre World War I and Breakdown Liberalconstitutional tradition Outco me QCA- Outcome Countries 0 0 C Czechoslovakia, Finland, Ireland 0 1 Austria, Estonia, Germany, Italy, Latvia, Lithuania, Poland, Spain, Greece, Portugal, Argentina, Cuba, Uruguay, Yugoslavia Australia, Belgium, Canada, Costa Rica, Denmark, France, Netherlands, New Zealand, rway, Sweden, Switzerland, United Kingdom, USA Source: Polity IV 2009 and own coding of degree of civil liberties. Coding: 1: Liberal-Constitutional Tradition before World War I; 0: Otherwise Stress of the Political System Based on David Eastons (1965) system theory it seems likely that political systems under extraordinary stress experience a breakdown, because the citizens are not willing to support a regime that does not satisfy their demands. And if the will of the people to support the political system decreases, the stress increases further. Especially the correlation between the absence of economic stress (that means economic growth, absence of an economic crises and low inflation) and the stability of democracy is well established (Przeworski et al. 1996, 2000). Stress for a political system can result from very different sources. The following factors are taken into account for the interwar democracies: a strong impact of a postwar crisis, a strong impact of the world economic crisis, a double crisis (that means a stong effect of a postwar crisis and the world economic crisis), a huge territorial loss and/or division of the previous state which led to a fragmentation in different states and last but not least as a measure of antisystem activities the strength of antisystem parties. I start with checking the presence of a strong postwar crisis. World War I was a traumatic experience for all countries taking part in that war, especially for the European countries. A typical example for a postwar crisis is Germany. In 1919 Free Corps took Berlin, the Spartakus leaders Karl Liebknecht and Rosa Luxemburg were murdered; soon afterward the German Revolution crushed. I followed the coding rule

15 15 by Berg-Schlosser/de Meur 1994 to distinguish between countries that experienced a post war crisis and those that did not. If we exclude the electoral oligarchies or not a postwar crisis is neither a necessary nor a sufficient condition for the breakdown of an electoral regime. The absence of a postwar crisis is also neither a sufficient nor a necessary condition for the survival of an electoral regime. Even if some of our Latin American cases, namely Costa Rica an Cuba were participants of World War I we find all our cases with a strong postwar crisis in Europe, especially in Central and Eastern Europe. It seems to make sense that a sudden drop in economic wealth endangers the persistence of an electoral regime and it is often said that indeed one of the major reasons why so many democracies failed in the interwar years was the strong effect of the World economic crisis. By using the data of Angus Maddison (2009) I test if there was a sharp drop in the economic performance after 1929 or not. All electoral regimes that broke down before 1929 are coded with 0, because the condition did not precede the outcome the economic crisis could not have played a role in these breakdowns. Table 9: Postwar Crisis and Breakdown Postwar Outcome QCA- Countries crisis Outcome Australia, Belgium, Canada, Costa Rica, Denmark, France, Ireland, Netherlands, New Zealand, rway, Sweden, Switzerland, United Kingdom, USA 0 1 Spain, Greece, Portugal, Argentina, Cuba, Uruguay, Yugoslavia 1 0 C Czechoslovakia, Finland 1 1 Austria, Germany, Italy, Estonia, Latvia, Lithuania, Poland Source for data: Berg-Schlosser/de Meur 1994 and own coding. Coding: 1: strong post-war crisis; 0: Otherwise. Table 10: Economic Crisis and Breakdown economic Outcome QCA- Countries crisis Outcome Italy, Lithuania, Poland, Spain, Portugal, Argentina, Cuba, Yugoslavia 1 0 C Australia, Belgium, Canada, Costa Rica, Czechoslovakia, Denmark, Finland, France, Ireland, Netherlands, New Zealand, rway, Sweden, Switzerland, United Kingdom, USA 1 1 Austria, Estonia, Germany, Latvia, Greece, Uruguay Source: Berg-Schlosser/de Meur 1994 and own coding based on Maddison Coding: 1: strong impact of the world economic crisis before the electoral regime broke down; 0: Otherwise.

16 16 Nearly all electoral regimes of the late 1920s were strongly affected by the world economic crisis of the late 1920s and early 1930s (for details see Saalfeld 2003, Zimmermann/Saalfeld 1988). The comparison shows that all of the surviving electoral regimes suffered from a strong impact of the world crisis. 6 Due to the fact that the fascists already took over the power in Italy in 1923 and seven more electoral regimes broke down before the Great Depression it is also obvious that a strong impact of the world economic crisis can not be a necessary condition for the breakdown of an electoral regime or a fascist takeover. So the often cited strong impact of the World economic crisis has not very much explanatory power on its own (referred to Weimar Germany see Lepsius 1978: 50). The next step is to clarify whether a combined strong effect of a postwar crisis and a strong effect of the Great Depression helps us a little more explaining the breakdown of hybrid and liberal-constitutional electoral regimes. Only cases which experienced both crises are coded with 1. The analysis shows that the impact of the double crisis (Arends/Kümmel 2000) does not help us to explain as much as one might think when he takes a look at the German case. Of the 14 failed electoral regimes only four suffered from a double crisis. It is also not very probable that the failures of the electoral regimes in Germany, Austria, Estonia and Latvia were related to the fact that the impact of world economic crisis ( ) was already the second profound crisis in less than ten years. For example in Germany the economy recovered rather quickly after the disastrous inflation of 1923 and therefore could not have had any direct impact on the democratic breakdown in Germany. During the golden years of Weimar between 1924 and 1929 the GNP per capita increased at an average rate of four per cent. Also two of the surviving electoral regimes Czechoslovakia and Finland experienced a strong impact of the World economic crisis and a strong post-war crisis. On the other hand Spain, Portugal, Yugoslavia, Cuba and Argentina neither experienced a crisis after World War I nor were they strongly affected by the world economic crisis before the electoral regime broke down. Table 11: Double crisis and Breakdown double Outcome QCA- Countries 6 The data gives us even the weird answer that a strong effect of the World economic crisis is a necessary reason for the survival of an electoral regime and the absence of a strong impact of the world economic crisis is a sufficient reason for the breakdown of an electoral regime. The second result is an artefact of our coding.

17 17 crisis Outcome 0 0 C Australia, Belgium, Canada, Costa Rica, Denmark, France, Ireland, Netherlands, New Zealand, rway, Sweden, Switzerland, United Kingdom, USA, 0 1 Italy, Lithuania, Poland, Spain, Greece, Portugal, Argentina, Cuba, Uruguay, Yugoslavia 1 0 C Czechoslovakia, Finland, 1 1 Austria, Estonia, Germany, Latvia Coding: 1 if there was a strong effect of a postwar crisis and the Great Drepression; 0: Otherwise. For each nation it means great stress when it loses much of its territory and/or many of its native people. After World War I there were huge territorial shifts in Europe. Three Empires the German Empire, the Austrian-Hungarian Empire and the Ottoman Empire perished from the map. The loss of Germany was extraordinary huge in Germany which lost 13 per cent of the previous territory and 15 per cent of the cultivated rural land (Arends/Kümmel 2000: 199). However, only two of our observed electoral regimes Germany and Austria suffered from a huge territorial loss and both broke down. Consequently territorial loss is a sufficient condition for the breakdown of an electoral regime. But the coverage is very low. Still huge territorial losses might be a good explanation why relatively well-developed countries turned into autocracies in the interwar years. In might as well explain the strengthening of fascist parties in some countries. In Germany the Nazis took over and in Austria the fascists took part in government. The absence of a territorial loss is also a necessary condition for the survival of an electoral regime. electoral regime survived with a huge territorial loss. Nevertheless, the condition is far from being also a sufficient explanation why an electoral regime survived the interwar years. Table 12: Territorial Loss and Breakdown Territorial Outcome QCA- Countries loss Outcome 0 0 C Australia, Belgium, Canada, Costa Rica, Czechoslovakia, Denmark, Finland, France, Ireland, Netherlands, New Zealand, rway, Sweden, Switzerland, United Kingdom, USA 0 1 Estonia, Italy, Latvia, Lithuania, Poland, Spain, Greece, Portugal, Argentina, Cuba, Uruguay, Yugoslavia Austria, Germany Coding: 1: territorial loss above 10 percent and/or division of the previous state which led to a fragmentation in different states; 0: Otherwise. Somehow surprisingly our stress factors, with the exception of a huge territorial loss, are neither sufficient nor necessary conditions for the survival or breakdown of an electoral regime. Still one should be very careful to say that stress does not matter at

18 18 all. It can be showed easily that all democracies at the point of breakdown had been in some kind of crisis (Linz/Stepan 1978). Last but not least, I am going to check in this part, how many voters prefered a different political system and did not support the electoral regime at all. There are good reasons to claim that usually for a democracy as for every political system it is enough if the people just follow by custom or rational choice (Linz 1978). Only if many people vote for anti-democratic parties or act anti-democratic the electoral regime gets in danger. The larger the vote share for left-wing and right-wing antisystem parties the higher the risk of a polarization of the party system and the breakdown of the electoral regime (Capoccia 2002, 2005, Ieraci 1993, Powell 1986, Sani/Sartori 1983). I define those parties as anti-democratic which try to overthrow the electoral regime, like the fascist and communist parties in the interwar period. Strong antisystem parties are neither a necessary nor a sufficient condition for the breakdown of an electoral regime. In Greece, Poland, Argentina, Cuba, Uruguay, Yugoslavia, Portugal the electoral regimes broke down without strong anti-democratic parties and in Belgium, Czechoslovakia, France, Netherlands the democracy survived despite strong antisystem parties. Table 13: Antisystem Parties and Breakdown Antisystemparties Outcome QCA- Countries Outcome 0 0 C Australia, Canada, Costa Rica, Denmark, Finland, Ireland, New Zealand, rway, Sweden, Switzerland, United Kingdom, USA 0 1 Greece, Poland, Argentina, Cuba, Uruguay, Yugoslavia, Portugal 1 0 C Belgium, Czechoslovakia, France, Netherlands 1 1 Austria, Estonia, Germany, Italy, Latvia, Lithuania, Spain Sources: Berg-Schlosser/De Meur 1994: 266, 268 and own coding. Coding: 1: votes for left-wing and right-wing antisystem parties (ca. 1930) above 15 per cent; 0: Otherwise. 5. Institutional Approaches Institutional variables have a special appeal for researchers because they are the only conditions politicians can change without great effort. It is relatively easy to quickly modify the electoral system, whereas it is a long way to improve the socioeconomic conditions. I examine the effect of two institutional variables: 1. a strong president and 2. a proportional electoral system. I have chosen these factors according to the following arguments in the literature: Juan Linz (1994) and many

19 19 others argued that presidential democracies are more prone to democratic breakdown (see e.g. also Kailitz 2004, Valenzuela 2004). The point is that the president and the majority of the parliament are independent of each other and there is no constitutional resolution for a deadlock between these two institutions. 7 Ferdinand Hermens (1941) brought forward the argument that a proportional electoral system would not only lead to a high fragmentation of the party system and a low government s stability, but also in the end to a breakdown of democracy. The presence or absence of a proportional electoral system is neither a necessary nor a sufficient condition for democratic survival or breakdown. Only if we exclude the electoral oligarchies a proportional electoral system is very close to be a necessary condition. Table 14: Proportional Electoral System and Breakdown Proportional Outcome QCA- Countries system Outcome 0 0 C Australia, Canada, New Zealand, United Kingdom, USA, 0 1 Spain, Portugal, Cuba, Uruguay 1 0 C Belgium, Costa Rica, Czechoslovakia, Denmark, Finland, France, Ireland, Netherlands, rway, Sweden, Switzerland, 1 1 Austria, Estonia, Germany, Italy, Latvia, Lithuania, Poland, Greece, Argentina, Yugoslavia Sources: hlen 1969, 2005; Coding: 1: proportional elecotoral system; 0: Otherwise. Table 15: Strong President and Breakdown Strong Outcome QCA- Countries President Outcome 0 0 C Australia, Belgium, Canada, Czechoslovakia, Denmark, France, Ireland, Netherlands, New Zealand, rway, Sweden, Switzerland, United Kingdom 0 1 Estonia, Italy, Latvia, Lithuania, Greece 1 0 C Costa Rica, Finland, USA 1 1 Austria, Germany, Poland, Spain, Argentina, Cuba, Portugal, Uruguay, Yugoslavia Sources: Constitutions of the countries. Coding: 1: strong president (system is presidential, quasipresidential or semi-presidential); 0: Otherwise. The presence or absence of a strong president is also neither a necessary nor a sufficient condition for the breakdown or survival of an electoral regime in the interwar years. For example Finland which had the most similar institutional design (semipresidentialism and a proportional electoral system) to Weimar among all European 7 In the discussion on the constitution of Weimar most prominently Karl Dietrich Bracher pointed out that the dual legitimacy of two popularly elected bodies independent of each other the president of the state and the parliament undermined the stability of the governments and led to a presidential dictatorship (Bracher 1962). Until today many political scientists stress that semi-presidentialism was a main problem of Weimar (Rüb 1994, Skach 2005).

20 20 countries (for details see Endemann 1999) and had similar problems democracy survived. On the other hand the most similar situation to the Nazi takeover in Germany was the fascist takeover in Italy and Mussolini's rise to power occurred in a constitutional parliamentary monarchy. 6. Actor-centered approaches Gregory Luebbert (1991) concludes in his structural analysis of interwar Europe: One of the cardinal lessons of the story I have told is that leadership and meaningful choice played no role in the outcomes [that means: liberalism, fascism or social democracy] (306). I do not agree with that. Especially under conditions of deep political crisis, institutional structures become malleable and the space of maneuver for the main actors is broader than during normal times (Dobry 1986). In such situations the actors are able to make certain choices that can increase or decrease the probability of persistence and stability of a regime (Linz 1978: 4). It is a fundamental problem for a democracy if there are not enough people who are principally in favor of the idea of democracy to the idea of autocracy and there is no hegemony of a democratic political culture. But the political activists always play a crucial role (Dahl 1989: 264). Dankwart Rustow (1970), the godfather of the actorcentered approach to transitions from and to democracy stated that a democracy cannot be stable without the decision of the main actors to sustain it. I first take a closer look at the question whether anti-democratic main actors intervened to bring the electoral regime down using a coup or a different method. It is obvious that this condition is in opposition to the socioeconomic conditions for instance very close to the outcome and it would be a great surprise if this condition was neither sufficient nor necessary. Indeed the behavior and action of the elites towards the electoral regime corresponds perfectly with the outcome. It is the rare case of a necessary and sufficient condition as well for the breakdown of an electoral regime (elites act anti-democratic) as for its survival (elites do not act anti-democratic). The necessity and sufficiency of the condition is formally not trivial, because there is variance in the condition and the outcome. But it seems hard to imagine that a democracy could break down without the involvement of any kind of anti-democratic elites. There are also good reasons to code unlike Berg-Schlosser/De Meur 1994 the presence of an anti-democratic

21 21 elite as well at least in three of our surviving democracies Belgium, Czechoslovakia, and Finland because of their strong antisystem parties backed by parts of the elite. If one would code the cases like this still anti-democratic elites would be a necessary and sufficient condition for the survival of an electoral regime, but the absence of an anti-democratic elite would no longer be a necessary condition for the survival of an electoral regime. Table 16: Anti-democratic elites and Breakdown Antidemocratic Outcome QCA- Countries elites Outcome Australia, Belgium, Canada, Costa Rica, Czechoslovakia, Denmark, Finland, France, Ireland, Netherlands, New Zealand, rway, Sweden, Switzerland, United Kingdom, USA Austria, Estonia, Germany, Italy, Latvia, Lithuania, Poland, Spain, Greece, Portugal, Argentina, Cuba, Honduras, Uruguay, Yugoslavia Source: Berg-Schlosser/De Meur 1994 and own coding. Coding: 1: active anti-democratic, 0: Otherwise. I turn now to the question if the actions of a leading politician mattered for the survival or breakdown of an electoral regime. First, I code if the head of state (president, prime minister or king) intervened in favor of democracy or not. Second, I code if a leading politician (president, prime minister or king) intervened in favor of autocracy. Such an action could have been a coup by leading politicians themselves or the acceptance of a legal revolution by the leading politicians. All cases in which there was no fundamental crisis caused by extremist actors are coded 0. An autocratic behavior of a leading politician is a sufficient condition for the breakdown of an electoral regime. Therefore, the absence of action of a leading politician to save democracy is a necessary condition for a breakdown, too. Every time a leading politician acted in favor of autocracy the democracy broke down and every time a leading politician acted in a time of crisis in favor of democracy like in Belgium, Czechoslovakia, and Finland (for the datails see Capoccia 2005). Under unfavorable conditions it seems to have a fundamental impact, if a leading politician supports democracy as in Finland and Czechoslovakia or whether he is not willing to defend it as Hindenburg in Weimar Germany or King Victor Emmanuel III of Italy. If we turn the tables it was a necessary condition for an electoral regime to survive the Interwar years that no leading politician acted in favor of an autocracy. Moreover,