Dominance and Vulnerability:

Transcription

1 ECPR Granada, April 25 Workshop on "Dominant parties and democracy" Draft and Vulnerability: The Structure of Party Competition and Democracy under the French Fifth Republic. Nicolas Sauger 1 Abstract: The French party system has experienced changes in its structures, from the hegemony of the Gaullist party in the 196s to a far more fragmented though quasi-bipartisan structure today. However, the structure of party competition in legislative constituencies is primarily characterised by stability throughout the Fifth Republic. In about one constituency out of four, a dominant party exists in 1962 just as 22. The impact of this limited case of local electoral dominance has nevertheless no straightforward consequences on the quality of the French democracy. If dominance is problematic because it contradicts the principle of the vulnerability of incumbents, it has no effect on the general level of turnout. Even if voters are sensitive to the structure of party competition, the impact of dominance on turnout is limited to the second round of the legislative elections. 1 CEVIPOF FNSP 98 rue de l Université F-757 Paris nicolas.sauger@sciences-po.fr - 1 -

2 The issue of one party domination has suddenly arisen again in France after the foundation of the Union for a Popular Movement (UMP) in 22. This new right-wing party, created by merging the former RPR with part of the UDF 2, had gained control on most of the different institutions of the Fifth Republic. President Chirac, around who was created the UMP, was elected in 22 with more than 8% of the votes (in the second round of the presidential election). The following legislative elections gave a majority to the UMP in the National Assembly (the UMP garners 365 out of 577 seats) 3. A majority of members of the Constitutional Council were nominated by UMP members 4 and a majority of local governments (regions, departments and cities) were controlled 5 by UMP representatives. Moreover, for the very first time under the Fifth Republic, a single party has had a majority in the Senate 6. Such a parliamentary and executive dominance of a single party was never achieved before. The congruence of the majorities among the executive (presidential majority), the National Assembly and the Senate is thus an exceptional feature (see table 1 below). Of course, French bicameralism is not symmetrical. This means that the congruence between the National Assembly and the Senate could be considered as not so remarkable because it has no huge impact on policy making (and even no impact on government stability). It is however a significant characteristic in defining dominance because Senate has veto powers in some legislative areas and some significant amendment powers (Tsebelis & Money, 1997), and because the President of the Senate has a right to nominate one third of the Constitutional Council. The lack of congruence between the National Assembly and the Senate did in fact cause political troubles especially for President de Gaulle and Mitterrand. The domination of the UMP over the French party system can however be questioned. Its parliamentary and executive dominance is in fact not based on electoral domination 7. In the first round of the presidential election, President Chirac only scored about 2% of the votes. The UMP furthermore suffered a severe defeat in the local and European elections of 24. The most symbolic aspect of this defeat was the swing of about half of the French regions to the left (2 2 In fact, the merger was even larger. It encompassed also the Liberal Democracy (which came out of the UDF in 1998) and part of the RPF (Pasqua's Eurosceptic party). See Haegel, 22 and Sauger, 23b. 3 The government is furthermore quasi mono-partisan since only one member of the UDF (G. de Robien) participates in. 4 The members of the Constitutional Council are nominated by the President of the Republic (Chirac since 1995), the President of the Senate (Christian Poncelet, UMP member, since 1998) and the President of the National Assembly. 5 The UMP lost this majority of local government in the 24 local and European elections, which gave control to the socialists of almost all the regions. 6 This majority was the result of the merger of two groups (RPR and Independent Republicans, which gathered members of either UDF or Liberal Democracy). However, the UMP majority has been now transformed in plurality since the results of the 24 senatorial elections. 7 The UMP is besides deeply divided by the rivalry of Jacques Chirac and Nicolas Sarkozy, currently president of the UMP

3 Table 1: congruence of majorities under the Fifth Republic President de Gaulle (gaullist) Largest group in AN Majority includes largest group? Number of seats of the group party in AN (in %) Number of effective parties in AN Largest group in Senate Majority includes largest group? Number of seats of the largest group in Senate (in %) Number of effective parties in Senate Yes 36% 4,8 Yes 23% 5,9 Yes 48% 3, Yes 41% 3, Gaullist 197 Yes 6% 2,5 Pompidou 1971 (gaullist) Yes 37% 4, Giscard d'estaing (UDF/PR) Mitterrand (PS) Chirac (RPR/UMP) PS RPR (Gaullist) Yes 31% 4,1 Independent Republican (center-right) PS Yes 58% 2,5 Centrist Union (UDF-CDS) No 37% 3,8 Yes 48% 3,2 Yes 44% 2,9 PS Yes 43% 3,4 UMP Yes 63% 2,2 RPR Yes 24% 6,1 Yes 22% 5,9 Yes 19% 6,9 Yes 21% 6,9 Yes 2% 6,9 No 21% 6,8 No 23% 6,2 Yes 22% 5,9 Yes 24% 5,2 Yes 28% 5 Yes 28% 5,1 Yes 29% 4,9 Yes 31% 4,7 Yes 3% 4,8 UMP Yes 47% 3,1-3 -

4 out of 22 regions being now led by a socialist). Thus, the domination of the UMP cannot be compared with the domination of the Gaullist party in the 196s 8. But even the Gaullist hegemony matches only the less restrictive definitions of party system domination 9 since it did not last very long (from 1962 to 1973 at most) and since the Gaullist have represented a majority in the National Assembly only from 1968 to Hence, most of the debates on party domination in France are in fact debates on bipartism (Duverger and Fauvet, 1965; Duverger and Goguel, 1967; Charlot, 197; Duhamel, 1997, Duhamel and Grunberg, 21 & 22) 11. Speaking about party system domination and democracy in France is however not meaningless, especially if the emphasis is put on the local / constituency level 12. Following the hypotheses proposed by Schlesinger and Schlesinger (199), patterns of party domination in the constituencies can be depicted. Hence, the issue raised in this paper is the evolution of party domination at the constituency level and its consequences on democracy and especially on turnout. The argument proceeds in three steps. Firstly, the paper proposes to link dominance and democracy through the concepts of vulnerability and accountability. Secondly, the evolution of party domination in French constituencies from 1958 is depicted. Thirdly, the consequences of the structure of competition on turnout are assessed. I., vulnerability, and accountability The literature on party dominance provides numerous definitions of the concept of dominance. As it was suggested in introduction, dominance is however generally defined in terms of congruence of majorities. Following Pempel (199), this involves a party being dominant in terms of numerical vote strength, in terms of parliamentary bargaining position and in terms of government (and possibly presidential) control. The issue of time is probably more debated. On the one hand, many authors propose to define dominance as some features of a party system at some point in the time. On the other hand, dominance can be viewed as the enduring ascendancy 8 Current electoral trends make the electoral domination of the UMP be not likely. Nevertheless, the popularity of Nicolas Sarkozy might contradict this prospect. 9 See Bogaards, 24, for a review of classical definitions. 1 Many analyses are thus quite ambiguous on the case of France. See for instance McGann 2 who exemplifies dominance with the case of France but who doesn't include France in his empirical investigations. 11 In this context, the changes in the structure of the French party system date back to the late 199s with the breakup of the UDF. If 1981 embodied the actual domination of the Parti socialiste on the left (Martin, 2), the break-up of the UDF in 1998 made it possible that the RPR definitively asserts its domination on the right side of the political spectrum (Sauger, 23a). 12 Even in terms of policy making, dominance is maybe not the best way of investigation despite the actual executive dominance of the UMP. For instance, the reform of the electoral system decided in 22 for regional elections is explicitly biased against the smaller parties (especially the National Front). But the decision that led to this electoral system is best depicted in terms of "cartelisation" since UMP and PS de facto collaborated on this project

5 of a party over a party system. In others terms, a truly dominant party is a party being able to maintain this position over a long period of time 13. Both approaches can be viewed as partly inefficient. If dominance is assessed through only one election, the analysis is likely to be biased by conjuncture effects. For instance, the UMP should be thus viewed as a dominant party in spite of the fact it is not really dominant in electoral terms. If dominance is assessed from a long period of time, different problems arise. When does the dominance of a dominant party start? Right from the beginning of the period? This would mean that a party which matches the criteria of dominance for one election but lose the following elections is a potential dominant party, which has failed. For instance, if the UMP wins the elections in 27 and 212, does this means that the UMP is currently a dominant party? A second answer to the initial question would be that dominance is achieved after a party has exerted power for a significant period. The idea is that electoral dominance allows the hegemonic party to change the rules of the electoral game to its own benefit, either by changing laws, creating or destroying institutions, or even by gradually changing the political culture (Pempel, 199). This, in turn, ensures future electoral success. In this case, dominance cannot be thought as a dichotomy. is gradually developed. In fact, part of the problem stems from the fact that dominance is often considered not only as a party system characteristic but also as a characteristic of the more general political system. Building on these different definitions of party dominance, we propose to define dominance as the probability a party has to keep on controlling government after the next general election 14. The more likely it is, the more dominant the party is. is thus here defined as belonging to the same dimension of party system characteristics as the "vulnerability of incumbents" 15 (Bartolini, 2), vulnerability and dominance representing the opposite poles of this dimension 16. Bartolini (1999 & 2) depicts vulnerability as one of the four key dimensions of a democratic electoral competition (the three other dimensions are the contestability of elections, the electoral availability and the decidability of the political offer). The vulnerability of incumbents has in fact two effects. On the one hand, vulnerability determines the room for manoeuvre of the party 13 This period can be defined as three consecutive elections (Sartori, 1976), twenty years (Blondel, 1968), "substantial" (Pempel, 199) or by that the fact that a party wins "usually" in the period analysed. 14 This can be also interpreted in terms of winning elections. 15 The idea of vulnerability has been conceptualised in a variety of ways. It refers to the same idea as the "closeness of the electoral outcome", the "uncertainty of the electoral result" or the "decisiveness of elections". It differs from classical studies of incumbency since it stresses more the potential turnover than its actual aspects. 16 Bartolini (2 : 52) defines vulnerability as "the possibility for an incumbent government to be ousted and replaced or otherwise modified in its composition as a result of changes in voters choices"

6 controlling government. Its electoral dominance makes it possible to loosen the constraint of the electoral demand 17 whereas its vulnerability forces it to be more responsive to citizen preferences 18. On the other hand, vulnerability is also a key issue for voters since it determines the potential impact of their vote on the electoral outcome. Vulnerability and dominance thus determine the existence of mechanisms of accountability, which are central in the definition of democracy (Manin and al., 1999 and especially Fearon, 1999). This has been particularly sketched through agency theory (Mitchell, 2; Strøm and al., 23). Elections are processes of delegation, from many principals (voters) to some agents or representatives (MPs,...) 19. Each process of delegation suffers from agency loss. Agency loss occurs when agents take action that is different from what the principal would have done. It results from the divergence of preferences among principals and agents as well as asymmetric information. Both phenomena produce risks of adverse selection (principal does not know the actual nature of its agent) and moral hazard (if principal lacks information about the agent's action, then the agent does not benefit from choosing any action other than his own preferences). Accountability refers to mechanisms by which agency loss may be contained. Fearon describes these mechanisms as "agency relations, in which one party is understood to be an agent who makes some choices on behalf of a principal who has powers to sanction or reward the agent" (Fearon, 1999: 55). Accountability therefore implies that principals have a capacity to impose sanctions to agents. Among the three main forms of sanction, which are the ability to block or amend decisions made by the agent, to "deauthorize" the agent (to remove him from office), and to impose specific penalties (Strøm, 23: 62), electoral sanctions are the more widespread. If an agent enjoys a dominant position, the probability of deselection through electoral sanction is very low. This results from the limited capacity of the electorate to change its vote in the short term (in economic terms, the electorate is not fully responsive to the degradation in quality of the product). Because electoral dominance prevents one of the main instruments of accountability from actually working, it can be thus defined as a break on the democratic process. 17 It also depends on electoral availability, i.e. the reactivity or elasticity of the electorate towards governmental decisions. 18 Unless the party is sure to lose the next elections. In this case, there is no incentives to be more responsive for the incumbent party. 19 Democracy is thus formed by a chain of delegation (from voters to MPs, from MPs to government, from government to ministers and from ministers to civil servants)

7 The following analysis is focused on the first step of the chain of delegation, the election of députés by French voters. The presidential race is not taken into account since patterns of domination (or absence of domination) are already well known. The emphasis is also put on the constituency level of analysis, the French electoral system being characterised by a single member two-ballot majority electoral system under the Fifth Republic 2. The issue of accountability is thus here understood as the accountability of each representative, who are also considered as party members 21. The superposition of the individual and partisan logics may seem confusing. It however only reflects the double nature of the French electoral system. It also enables to take into account the significant role of incumbent députés in the French parties. II. Patterns of party competition in France: The principal dynamic of the French party system under the Fifth Republic has been that of the so-called "bipolar quadrille". By the end of the 197s, four parties of approximately equal strength were monopolising over 9 per cent of the vote in their respective left and right blocs (Parodi, 1991). Nevertheless, this end-state had taken twenty years to produce, concluding in 1978 with the formation of the Union pour la Démocratie Française (UDF). And the quadrille did not last very long. It had already been shattered by the decline of the French communist party and the rise of the National Front in the early 198s (Martin, 2). It definitively came to an end in 1998 with the break-up of the UDF. The fragmentation of the French party system in the 198s and 9s has however been partly balanced by the progressive confirmation of the leading status of one rightwing (the RPR and then UMP) and one leftwing (PS) parties. Thus, if one-party domination has characterised the early Fifth Republic, more competitive patterns of party competition emerging from 197s. The fragmentation of the party system in the 198s and the 199s has thereafter even reinforce the competitiveness of the party system, the vulnerability of the two leading parties being best exemplified by Lionel Jospin's defeat in the first round of the 22 presidential election. This classical interpretation of the dynamics of the French party system can be challenged through few simple indicators. Six of these indicators are charted in graphs 1 to 3. The two first graphs deal with indicators of fragmentation through the concept of the effective number of 2 The only exception is the election of 1986 (with a proportional electoral system). This election will be generally excluded from our analysis. 21 Even if the incumbent retires, the logics of accountability is transferred to the new candidate of the same party

8 Graph 1: fragmentation of the French party system (legislative elections ) Number of effective parliamentary parties Number of effective electoral parties Number of effective electoral parties per constituency Graph 2: fragmentation of the French party system (presidential elections ) Number of effective presidential candidates Graph 3: incumbency and majority elected députés in French legislative elections (1986 elections excepted),9,8,7,6,5 Proportion of incumbents,4,3,2, Proportion of députés elected in the first round

9 parties 22. It is assessed for both legislative and presidential elections. For legislative elections, we calculate the effective number of national electoral parties, of national parliamentary parties and the mean of the effective number of electoral parties throughout the constituencies 23. The third graph charts the evolution of the proportion of incumbents among députés and the evolution of the proportion of députés elected in the first round 24. If the history of competition for the presidential race more or less matches with the evolution depicted, legislative elections are chiefly characterised by stability in the structure of party competition. This is especially obvious if competition in the constituencies is taken into account. The effective number of parties varies only from 3 to 4. And these variations are partly due to electoral swings between right and left, the left electoral progress causing the decrease of the effective number of parties 25. The French party system thus displays characteristics of loose national - constituency linkage at the electoral level. Cox's "index of inflation" 26 from the constituency to the national levels has been equal to.325 on average since 1958 (see table 2, below), i.e. equivalent to India's level (Cox, 1999: 157). The evolution of the vulnerability of French governments (from a situation of absence in alternation before 1981 to a situation of systematic alternation after 1981) does not therefore reflect the evolution of vulnerability in constituencies. Since this can also be viewed as a significant condition of democracy (see part I above), the following analysis will be focussed on the assessment of vulnerability in constituencies. Table 2: Cox's inflation index (electoral coordination between constituency and national levels). Cox's inflation index ,43,37,31,3,37,24,3,25,31,37,36,23 22 The effective number of parties is based on a measure of the probability that any two voters randomly chosen from the electorate have voted for different parties (see Lakso and Taagepera, 1979). N= 1 1 (if the total of votes cast is large enough) 2 c vij vij 1 c vij = = = = c c c i 1 vij vij 1 i 1 vij i 1 i 1 = i 1 with v ij : votes obtained by party i in the constituency j; c : party's number (parties possibly being candidates). 23 All the results are new calculations from official sources. 24 Députés can be elected in the first round of the election if they gather a majority of votes cast. 25 This can be explained by partisan bias against leftwing parties in the electoral system (see King and Gelman, 1991) and also by the politics of electoral coalitions. 26 This index is simply based the difference between the effective number of parties at the national level and the mean of the effective number of parties in constituencies (Cox, 1999). Nnat Nconst I= N nat - 9 -

10 The concept of vulnerability is closely link to the idea of the uncertainty of the electoral outcome. It "refers to the psychological effect linked to the absence of safety, rather than the actual result [...]; closeness and uncertainty may not result in turnover but still provide their supposed effect on competition " (Bartolini, 2: 52). A great variety of empirical measures have been proposed to assess this kind of uncertainty. On the one hand, different indexes have been proposed. On the other hand, a method of visualisation of party strengths ("Nagayama triangle") has also been recently applied, especially to Italy and Japan (Reed, 21; Grofman & al., 24; Taagepera, 24). "Nagayama triangles" (and more especially the segmented version of Nagayama triangles) are used here as an exploratory method. Nagayama triangles are based on constituency level data. This kind of diagram is used to show the vote share of the largest party on the x-axis, and the vote share of second largest party on the y-axis. Because the second largest party must receive fewer votes than the largest party, the feasible set of values in the diagram lies within a triangle bounded by the x-axis and segments of the line x-y= and x+y=1 27. These diagrams thus display information about the relative score of the two largest parties and information about the aggregated score of all the other parties (it is the distance between the plot and the nearest segment). We have also added two other segments inside the triangle 28. On the left side of the triangle, any plot inside the two segments means that the difference of scores between the two largest parties is below a fixed limit (here two standard deviations of the score of the largest party); on the right side of the triangle, any plot between the two segments means that the two leading parties represent a total of 85% of the votes in the constituency. Graphs 3 to 14 are Nagayama diagrams for the first and second largest parties; Graphs 15 to 26 are Nagayama diagrams for the first and third largest parties. [see Nagayama file appended] First of all, these diagrams show the stability of party competition in constituencies over more than forty years. Competition has never been strictly bipartisan (two party of roughly equal strength representing more than 85% of the votes). The differences among the three leading candidates in each constituency have moreover been significant most of the time (the difference between the two leading candidate being roughly equal to two standard deviation of the score of the leading party in the constituency). Closeness of competition in most of the constituencies has been achieved only in 1981 and 1997 (times of electoral victories of the left). Moreover, the 27 This triangle can be translated to represent the vote share of the third largest party. In this case, the segments are defined by x-y= and x+2y=1. 28 The position of these segments (x-y=,15 and x+y=,85) have been chosen according to the standard deviation of the score of the largest parties in the constituencies (,7 on average). This measure has been repeated three times since significant redistricting occurred in 1966 and The average of the standard variations across districts does not significantly vary across time

11 dispersion of the plots is higher when rightwing parties win with a large margin (especially 1968 and 1993). This highlights a pattern of local electoral dominance with bias towards the right. These results can be more systematically tested with rough figures. Table 3 proposes fifteen variables summarizing Nagayama diagrams 29. Despite significant "blips", the overall picture is that of stability. However, it makes it possible to notice the impact of the broad national evolutions at the local level (the partial unification of rightwing parties into the UMP has led for instance to widen the gap between the second and third parties). But in no case have patterns of local electoral domination been dramatically altered throughout the period. In about one third of the constituencies, the largest party benefits from a significant advance on the second largest party (more than two times the average standard deviation of the score of the largest party). In fact, all these results largely confirm the theoretical hypotheses of G. Cox on the effect on uninominal two-ballot majority electoral systems (Cox, 1997). Table 3: the structure of competition in French constituencies ( ) (all figures are % of votes or % of constituencies) Mean V 1 36, ,9 46,7 38,9 39,1 46,8 44,8 38,5 34,1 39,7 Mean V 2 24,6 26,4 26,9 25,8 24,8 25,3 32,4 31,9 21,3 25,5 26,3 Mean V 3 17, ,2 15,1 17,1 17,9 14,8 13,2 13,9 15,8 13,2 V 1 > 5% 9,2 2,6 15,1 32,4 1,3 13,5 33,1 2,9 13 1,3 9,7 V 1 < 25% 9,5 1,5 2,3 1,9 8,1 2,5,7 V 2 + V 3 > 5% 12 21,5 25,6 2 12,8 11,6 36,3 17,5,4 3,4 3,2 V 3 < 25% 96,3 92,5 91,6 92, ,1 91,4 97,1 1 99,6 99,1 V 1 + V 2 > 85% 4,7 9,2 4 1,7 1,3 2, ,6 V 1 + V 2 ]65%, 85%] 27,3 51,6 66,4 71,2 45,8 41, ,9 29,7 24,1 56,8 V 1 - V 2 < 7% 46,2 31,8 29,6 21,4 28, ,5 34,2 27,6 51,7 33,3 V 1 - V 2 ]7%, 15%] 26, ,6 2,6 31,5 27,2 32,5 3,1 23,4 31,4 29,4 V 1 V 3 < 7% 16,1 9,2 4,8 2,9 7,4 11,4 3 2,3 8,5 1,8 2,9 V 1 V 3 ]7%, 15%] 29,9 27,7 14,9 1,5 2,2 24,7 11,2 7,9 18,6 29,7 11,7 V 2 V 3 < 7% 58,7 51,4 43,7 39,1 52, ,6 15,7 57,7 42,2 29 V 2 V 3 ]7%, 15%] 29,7 27,7 3,9 31,3 33,2 3 2,3 2,5 3,5 33,9 29,5 (V 1, V 2, V 3 : vote shares of the largest party, second largest party and third largest party in each constituency) 29 In fact, we also added a few indicators directly drawn from game theory analysis : they are especially the vote share of the third largest party and the total amount of votes gathered by the second and third largest party. In terms of game theory, "a dominant party is a party such there is at least one pair of mutually exclusive losing coalitions, each of which the dominant party can join to make winning, but which cannot combine with each other in the absence of the dominant party to form a winning coalition" (Laver and Benoit, 23 : 22). It can be proved then that the largest party cannot be dominant if, the largest party scores less than 25% of the votes, if the third largest party gathers more than 25% of the votes or if the second and third largest party gather a majority of votes

12 A second approach of party dominance is the actual measure of dominance through some quantitative index. The empirical literature offers various operational definitions of electoral closeness. Among these various definitions, the most popular are measures of electoral margin (most simply v 1 v 2 ) and measures of competitiveness understood as the probability that any two voters have voted for a different party (fractionalisation index F = c i i i= 1 c c i i v i= 1 v v 1, v 1 i= 1 competitiveness index C= c c i c i= 1 v, or entropy index E= c i= 1 vi i= 1 v ln i i c c v = = i vi i 1 i 1 v 3, see for instance Endersby and al., 22; Kirchgassner and al., 1997). Measures of electoral margin are best adapted to the issue of electoral domination since they actually tackle the particular role of the largest party. For both theoretical and empirical reasons, the raw electoral margin (v 1 v 2 ) is not fully adapted to the French case. More especially, it does not take into account both the two-ballot majority system and the multiparty feature of the party system. Furthermore, the actual past record and present uncertainty have to be incorporated in the measure of vulnerability. That is why we propose a new measure of electoral dominance in constituencies: v1 σ D= v2+ v This measure relies on the comparison of the largest party and the second and third largest parties. This is drawn from the theoretical study of dominance through coalition theory (Laver and Benoit, 23). It is also closely related to Cox's analysis of uninominal two-ballot majority electoral systems (Cox, 1997). This measures also includes the standard deviation of the vote share of the largest party in the constituency to incorporate the dimension of time. A party is here viewed as dominant if its vote share minus the standard deviation of the vote share of the largest party in the constituency over the period of time studied is superior to the addition of the scores of the second and third largest parties (i.e. D > 1). The measure is robust if we consider its predictive power. Tables 4 and 5 show that almost all dominant parties in the first round of the election win the election (at the first or second round) 31. Besides, the measure also reflects dominance on time since dominant parties have a far higher ( v1 ) 3. 3 v i : votes obtained by party i ; c : party's number (parties possibly being candidates). 31 We have also introduced the intermediary level or "marginal" constituencies as an intermediate level between strictly dominant constituencies and strictly competitive constituencies (see Stern, 1972; Meltz, 1973)

13 probability or re-election than non-dominant parties. We can also notice that dominant parties

14 Table 4: probability that the largest party wins (first or second round) according to level of dominance Dominant (D>1) ,3 99, ,7 97, Marginal (D ]1,,75]) 82,2 92, ,6 91,5 94,1 73,3 91,9 95,4 93,8 93,6 Competitive (D<,75) 63,4 74,9 45,1 52,7 72,8 7,2 58,8 72,4 7 67,3 71,4 Table 5: probability that the largest party at t wins at t x election t = 1988 t =1993 t =1997 t +1 in dominant constituencies 85,1 85,2 87,9 t +2 in dominant constituencies 81 74,9 - t +3 in dominant constituencies 83,7 - - t +1 in non dominant constituencies 47,3 49,9 65,4 t +2 in non dominant constituencies 7,4 67,1 - t +3 in non dominant constituencies 59,3 - - Tables 6 &7: correlation among levels of dominance over time ** : significant at.1 level * : significant at.5 level ,52**,51**,36**,52**,51**,5**,51**,51**,56**,36**,5**,56** ** : significant at.1 level * : significant at.5 level ,68**,5**,12*,4,68**,48**,9*,3,5**,48**,8 -,1,12*,9*,8,58**,4,3 -,1,58**

15 are significantly less dependant on electoral change (the large victory of right wing parties in 1993 altered the results chiefly in non dominant constituencies). The robustness of our indicator over time can be also proven by the spatial stability of dominance (see tables 6 & 7; tables are separated because of 1988 redistricting). The only exception is the election of It clearly reflects the episode of realignment of the French electorate that tool place at this time (and which was fostered by party change: creations of the RPR and UDF, common platform of the left) (Dupoirier & al., 1988; Martin, 2). Table 8 provides the measures of electoral dominance of the largest party in French constituencies from The results are congruent with previous conclusions. Patterns of dominance have not dramatically changed throughout the period. Constituencies with a dominant party represent about 24% of the constituencies on average. However, these measures show significant changes in the short term, depending on swings between left and right (the electoral defeat of the left makes dominance decrease, except in 1981, and large victories of the right, such as 1968 and 1993, makes dominance achieve its highest levels). Table 8: distribution of constituencies according to the level of electoral dominance of the largest party Dominant (D>1) 16,1 25,4 2,6 42,2 18,3 18,6 23,2 23,4 39,5 8,6 27,4 Marginal (D ]1,,75]) 19,8 25,4 28,8 25,6 24,8 25,3 36,7 39,6 19,6 2,2 31 Competitive (D<,75) 64,1 49,2 49,4 3,9 56,3 56,1 4,1 36,8 4,9 71,2 41,6 III. Party competition and turnout in France Beyond the issue of accountability, dominance and democracy are often linked through the question of the impact of dominance on turnout. There is a large body of literature debating both the theoretical reasons and the empirical validity of the idea that dominance has a negative impact on electoral participation (Blais, 2; Franklin, 22 and 24, Norris, 22). From the initial proposition of Riker and Odershook (1968), many studies have demonstrated the existence of a correlation between closeness of the race and turnout. For instance, Heath and Taylor (1999) have found that the difference in national share of the vote between the major parties is positively correlated with turnout

16 The link between turnout and electoral dominance is here used in two ways. Firstly, we test the hypothesis of the impact of dominance on turnout. Secondly, we use turnout to assess what is the subjective perception of dominance by voters, i.e. which indicator is the more closely linked with turnout. The relationship between electoral dominance and turnout is not straightforward. Table 9 displays coefficients of correlation (Pearson) between turnout and dominance for both first and second rounds of the legislative elections from Electoral dominance in the first round has no general impact on turnout, excepted four important cases (1967, 1988 and 1993 and 22). Moreover, in these exceptional cases, the coefficients are positive but weak, i.e. dominance makes turnout increase. Unfortunately, no explanation can be provided: this result may be an artefact or represent mobilization dynamics (however, the context of these four cases are quite different). What is probably more interesting is the significant correlations between dominance in the first round and turnout in the second round. This correlation is even stronger if the difference in turnout between both rounds is taken into account. Coefficients are also negative, in accordance with theoretical expectancies. However, these coefficients are weaker, and even not significant, in case of electoral victory of the left (1981, 1988, 1997). This can be explained by the specific patterns of party competition produced by electoral victories of the left (see II. above). Table 9: correlation between dominance and turnout at the constituency level Correlation between dominance in the fist round and turnout (first round) Correlation between dominance in the fist round and turnout (second round) ,5 -,4,18** -,2,8,1 -,4,19**,13**,7,1* -,19** -,15** -,29** -,32** -,6 -,17** -,7 -,3 -,12** -,2 -,2 Correlation between dominance in the first round and change in turnout between the two rounds -,2** -,23** -,38** -,38** -,13** -,19** -,1 -,9 -,12** -,7 -,11* ** : significant at.1 level * : significant at.5 level

17 To assess more precisely the impact of dominance on turnout, this impact has to be controlled by other variables. Unfortunately, until 1981, no control variable is available, except the number of registered voters by constituency. This variable has no impact on turnout neither in the first round nor in the second round. Regression analysis however reinforces the correlation analysis since dominance is systematically a significant variable of the regression in the second round (and never in the first round). From 1988, the impact of electoral dominance can be controlled by census data. Besides the number of registered voters in each constituency, six variables have been chosen in accordance with classical explanations of turnout: the percentage of the population between 18 and 25 years old, the percentage over 65 years old, the percentage of farmers, the percentages of manual workers, the percentage of population living in rural areas, the percentage of foreigners over 18 years old in the population of the constituency. The dependant variable is either turnout in the first round (t 1 ) or the difference of turnout between rounds (t 2 -t 1 ). The results of the estimations through OLS regressions are provided in table 1. The two first columns of each date largely confirm previous results. Turnout in the fist round is chiefly structured by socio-economic logics even though they explain only part of the variance (about one third of the variance, in average, the adjusted R² being equal to.283 to.465). plays no role in explaining turnout. In the second round, dominance has an impact on turnout. The impact of dominance is besides negative on turnout. It is however rather weak and hazardous. Despite the fact that some other variables are also more or less significant, no explanation in variation of turnout is really achieved. Adjusted R² scores are very low and much of the variance is not explained. In fact, turnout variation between rounds primarily depends on the vote share of the winning party in the second round (column 3). The larger the victory is, the lower turnout is. This result can be interpreted in to different ways. Firstly, we can suppose that voters have better expectations about the electoral outcome than what is offered by our index of dominance. In particular, voters actually take into account electoral coalitions and probability of electoral transfers. Hence, domination of one party in the first round is not a so good predictor of electoral victory (though we have shown quite good results in part II) and domination may be also defined as the potential attraction of votes, taking into account the issue of proximity among parties in the first round to predict results in the second round. This may also reflects the difference between domination and closeness of the race (domination does not primarily rely on coalition potential whereas closeness does). Secondly, we can also suppose that that the margin of victory depends on actual turnout through mechanisms of skewed turnout: the margin is all the

18 more larger the losing candidate fails to mobilize its potential electorate (because he is views as an expected loser or because of some other reasons)

19 Table 1: estimation of turnout in constituencies constant t 1 t 2 -t 1 t 2 -t 1 t 1 t 2 -t 1 t 2 -t 1 t 1 t 2 -t 1 t 2 -t 1 t 1 t 2 -t 1 t 2 -t 1,683*** (,27),13*** (,9),267*** (,54) -,77*** (,18) V 1 - -,349*** (,21),794*** (,26) -,9 (,73),337*** (,43),732*** (,26),15 (,48),353*** (,24),82*** (,26) -,19 (,31) -,3 (,3) -,3** (,11) -, (,5) -,34** (,11) -,1 (,4) -,18** (,6) - -,558*** (,11) - - -,634*** (,2) - - -,55*** (,12) - -,22*** (,23) -,352*** (,15) registered voters (*1-5 ),27* (,16) -,43 (,32),5 (,13) -,1 (,14),17 (,38) -,3 (,22) -,4 (,13) -,57 (,23) -,37** (,12) -,6 (,11) -,31 (,13) -,38*** (,9) Pop age [18-24] -,381*** (,94) -,336* (,19) -,72 (,75) -,54*** (,89),123 (,253),15 (,144) -,339*** (,88),112 (,163) -,88 (,81) -,393*** (,91),26 (,16) -,166 (,74) Pop age >65 -,24 (,46) -,63 (,89),96** (,35) -,265*** (,45),165 (,124),54 (,71) -,166*** (,44) -,71 (,82) -,1 (,41) -,17*** (,46) -,4 (,54) -,21 (,37) Farmers,431*** (,73),152 (,158) -,41 (,62),325*** (,69) -,128 (,211),23 (,12),249*** (,68),487*** (,128),159 (,64),523*** (,69),264** (,85),99 (,59) Workers -,15 (,34) -,414*** (,68) -,49* (,28) -,95** (,32),77 (,93),36 (,52),23 (,32) -,25*** (,58) -,85** (,29) -,47*** (,32) -,15 (,39),29 (,27) Rural population (<5h),3 (,27),168*** (,53),59** (,21),15 (,26),82 (,76),37 (,43),6 (,25),11 (,47),3 (,23),21 (,26),2 (,31) -,11 (,22) Foreigners -,29*** (,33) -,455*** (,61) -,57 (,25) -,223*** (,32) -,51 (,87) -,19 (,49) -,23*** (,32) -,75 (,58) -,4 (,29) R²,465,251,884,283,15,681,323,71,771,41,19,572 F 61,268*** 19,344*** 419,934*** 28,268*** 1, ,799*** 33,992 6,192*** 231,776 49,164*** 8,64*** 84,599*** V 1 : vote share of the largest party in the second round * : significant at.1 level ** : significant at.5 level *** : significant at.1 level -,182*** (,33) -,135*** (,38) -,145*** (,26)

20 Thus, electoral dominance in constituencies cannot be viewed as impairing the quality of the French democracy because of its impact on turnout. Is voters are sensitive to the structure of competition, this sensitivity is true only in the second round of elections. There is even no remaining effect of large electoral victories on following elections. The score of the largest party at time t 1 has no significant impact on turnout at time t 2 or further (in fact, the score of the largest party is weakly correlated with turnout but the variable is not significant in regressions). Conclusion This exploratory paper ends up at two main results. Despite significant changes in the French party system at the national level, the first result is that patterns of electoral dominance have existed throughout the Fifth Republic at the constituency level for legislative elections. The second result assessed mixed consequences of this electoral dominance. On the one hand, dominance has a significant impact on the quality of democracy because it impairs vulnerability of incumbents and hence decrease the constraint of accountability. However, this theoretical expectation cannot be easily empirically tested. On the other hand, dominance has an impact on turnout. But voters sensitivity to patterns of electoral competition is limited in time. It only affects turnout in the second round of legislative elections and thus cannot be viewed as problematic for democracy. References : Bartolini, Stefano, 1999, "Collusion, competition and democracy (I)", Journal of Theoretical Politics, 1999, 11 (4), pp Bartolini, Stefano, 2, "Collusion, competition and democracy (II)", Journal of Theoretical Politics, 2, 12 (1), pp Blais, André, 2, To vote or not to vote, Pittsburgh, University of Pittsburgh Press. Blondel, Jean, 1968, "Party systems and patterns of government in Western democracies", Canadian Journal of Political Science, 1 (2), pp Charlot, Jean, 197, "Du parti dominant", Projet, 48, pp Cox, Gary, 1997, Making votes count, Cambridge, Cambridge University Press. Cox, Gary, 1999, "Electoral rules and electoral coordination", Annual Review of Political Science, 2, pp Duhamel, Olivier and Gérard Grunberg, 21, "Système de partis et 5 es Républiques", Commentaire, 24, pp Duhamel, Olivier and Gérard Grunberg, 22, "Les partis et la 5e République : post-scriptum", Commentaire, 25, pp Duhamel, Olivier, 1997, "Derrière le brouillard, le bipartisme?", Etat de l'opinion 1997, pp Dupoirier, Elisabeth, Capdevielle, Jacques, Ysmal, Colette, Grunberg, Gérard, Schweisguth, Etienne, 1988, France de gauche, vote à droite, Paris, Presses de la FNSP

21 Duverger, Maurice and François Goguel (dirs), 1967, Permanence et changement dans le système de partis français, Paris, Association Française de Science Politique. Duverger, Maurice and Jacques Fauvet (dirs), 1965, Le bipartisme est-il possible en France?, Paris, Association Française de Science Politique. Endersby, J., Galatas, S., Rackaway, C., 22, "Closeness counts in Canada", Journal of Politics, 64 (2), pp Fearon, James, 1999, ""Electoral accountability and the control of politicians", in B. Manin, A. Przeworski, S. Stokes (dirs), Democracy, accountability and representation, Cambridge, Cambridge University Press, pp Franklin, Mark, 22, The dynamics of electoral participation, in L. LeDuc, R. Niemi, and P. Norris (dirs), Comparing democracies 2, Thousand Oaks, Sage, pp Franklin, Mark, 24, Voter turnout and the dynamics of electoral competition in establish democacies, New York, Cambridge University Press. Grofman, Bernard, Chiaramonte, A., D'Alimonte, R., and S. Feld, 24, "Comparing and contrasting the uses of two graphical tools for displaying patterns of multiparty competition", Party Politics, 1 (3), pp Haegel, Florence, 22, "L'UMP et la refondation des partis de droite", Revue Française de Science Politique, 52 (5-6), pp Heath, A., Taylor, B., 1999, New sources of abstention?, in G. Evans, P. Norris (dirs), Critical elections, London, Sage. King, Gary, Gelman, Andrew, 1991, "Systemic consequences of incumbency advantage in U.S. House elections", American Journal of Political Science, 35 (1), pp Kirchgassner, Meyer, Himmern, 1997, "Expected closeness and turnout", Public Choice, 91 (1), pp Lakso, Markku, Taagepera, Rein, "Effective number of parties: a measure with application to West Europe", Comparative Political Studies, 1979, 12, pp Laver, Michael, Benoit, Kenneth, 23, "The evolution of party systems between elections", American Journal of Political Science, 47 (2), pp Manin, Bernard, Adam Przeworski and Susan Stokes (dirs), 1999, Democracy, accountability and representation, Cambridge, Cambridge University Press. Martin, Pierre, 2, Comprendre les évolutions électorales. La théorie des réalignements revisitée. Paris : Presses de Sciences-Po. McGann, Anthony, 2, "Social choice and hegemony: party system dominance in Western Europe", paper presented at the APSA congress, Washington, 2. Meltz, D., 1973, "An index for measurement of interparty competition", Behavioral Science, 18, pp Mitchell, Paul, 2, "Voters and their representatives: electoral institutions and delegation in parlimentary democracies", European Journal of Political Research, 37, pp Norris, Pippa, 22, Democratic Phoenix, New York, Cambridge University Press. Parodi, Jean-Luc, 1991, "Le nouvel espace politique français", in Y. Mény (dir.), Idéologies, partis politiques et groupes sociaux, Paris, Presses de la FNSP, pp Pempel, T. J. (dir.), 199, Uncommon democracies: the one-party dominant regimes, Ithaca, Cornell University Press. Reed, Steven, 21, "Duverger's law is working in Italy", Comparative Political Studies, 34, pp Riker, William H., Ordeshook, Peter, 1968, A theory of the calculus of voting, american Political Science Review, 62, pp Sartori, Giovani, 1976, Parties and party systems, Cambridge, Cambridge University Press. Sauger, Nicolas, 23a, Les scissions de l'udf : unité et dissociation des partis, mécanismes de transformation de l'offre partisane, Paris, Thèse de l'iep de Paris. Sauger, Nicolas, 23b, " The UDF in the 199s: the Break-up of a Party Confederation", in J. Evans (éd.): The French party system, Manchester : Manchester University Press, 23, pp Schlesinger, Joseph, Schlesinger, Mildrer, 199, "The reaffirmation of a multiparty in France", American Political Science Review, 84 (4), pp Stern, M., 1972, "Measuring interparty competition", Journal of Politics, 34, pp Strøm, Kaare, 23, "Parliamentary democracy and delegation", in K. Strøm, W. Müller, T. Bergman, (dirs), Delegation and accountability in parliamentary democracies, Oxford, Oxford University Press, pp

22 Strøm, Kaare, Müller, Wolfgang, Bergman, Torbjörn (dirs), 23, Delegation and accountability in parliamentary democracies, Oxford, Oxford University Press. Taagepera, Rein, 24, "Extension of the Nagayama triangle for visualization of party strenghts", Party Politics, 1 (3), pp Tsebelis, George and Jeannette Money, 1997, Bicameralism, Cambridge : Cambridge University Press

23 Graphs 3-14 : scores of the first and second parties in the first round of legislative elections L1958_1 L1968_1,5,5,4,4,3,3,2,2,1,1,1,2,3,4,5,6,7,8,9 1,1,2,3,4,5,6,7,8,9 1 L1962_1 L1973_1,5,5,4,4,3,3,2,2,1,1,1,2,3,4,5,6,7,8,9 1,1,2,3,4,5,6,7,8,9 1 L1967_1 L1978_1,5,5,4,4,3,3,2,2,1,1,1,2,3,4,5,6,7,8,9 1,1,2,3,4,5,6,7,8,9 1

24 L1981_1 L1997_1,5,5,4,4,3,3,2,2,1,1,1,2,3,4,5,6,7,8,9 1,1,2,3,4,5,6,7,8,9 1 L1988_1 L22_1,5,5,4,4,3,3,2,2,1,1,1,2,3,4,5,6,7,8,9 1,1,2,3,4,5,6,7,8,9 1 L1993_1,5,4,3,2,1,1,2,3,4,5,6,7,8,9 1

25 Graphs : scores of the first and third parties in the first round of legislative elections L1958_1 L1968_1,35,3,25,2,15,1,5,1,2,3,4,5,6,7,8,9 1,35,3,25,2,15,1,5,1,2,3,4,5,6,7,8,9 1 L1962_1 L1973_1,35,3,25,2,15,1,5,1,2,3,4,5,6,7,8,9 1,35,3,25,2,15,1,5,1,2,3,4,5,6,7,8,9 1 L1967_1 L1978_1,35,3,25,2,15,1,5,1,2,3,4,5,6,7,8,9 1,35,3,25,2,15,1,5,1,2,3,4,5,6,7,8,9 1

26 L1981_1 L1997_1,35,3,25,2,15,1,5,1,2,3,4,5,6,7,8,9 1,35,3,25,2,15,1,5,1,2,3,4,5,6,7,8,9 1 L1988_1 L22_1,35,3,25,2,15,1,5,1,2,3,4,5,6,7,8,9 1,35,3,25,2,15,1,5,1,2,3,4,5,6,7,8,9 1 L1993_1,35,3,25,2,15,1,5,1,2,3,4,5,6,7,8,9 1