Party Systems Determinants of Electoral Reform in Post-communist States

Transcription

1 Draft: March 2009 Please do not cite without permission Party Systems Determinants of Electoral Reform in Post-communist States Jack Bielasiak and John W. Hulsey Department of Political Science Indiana University-Bloomington Bloomington, IN Paper prepared for the Workshop Why Electoral Reform? The Determinants, Policy and Politics of Changing Electoral Systems, European Consortium for Political Research Joint Sessions of Workshops, Lisbon, April This paper addresses the issue of how the structure of party systems affects the initiation and direction of electoral reforms in post-communist emerging democracies. The analysis examines the relationship between party systems, defined in terms of the effective number of political actors, electoral volatility, and shares of party votes and seats, and the presence of diverse reform types. The findings reveal a significant number of reforms that undermine the institutional inertia claim, although the frequency of reforms declines appreciably with successive electoral cycles. There is no unidirectional trend, as both permissive and restrictive reforms are in evidence. These findings do not support the theoretical assertions linking party system fragmentation or electoral volatility to a specific direction of electoral change.

2 The collapse of authoritarian and communist regimes during the third wave of democratization ushered in a new political era across the globe. The monopoly of power vested in single parties, military regimes, or authoritarian personalities was replaced by a legitimating principle that evolved, in most cases, towards a contested version of politics. For many, the change signaled the embrace of democratic principles associated with the politics of choice, including the design of electoral institutions to invigorate popular participation and political competition. At about the same time, several mature democracies shed long-term stability by introducing significant electoral changes, whether in Italy, Japan or New Zeeland. Both developments, the electoral reforms in the advanced industrial world and the design of electoral systems in nascent democracies, provided strong impetus for scholarly attention to questions of electoral engineering and reform (Lijphart, 1994; Farrell, 2001; Birch, 2003; Norris, 2004; Colomer, 2005; Diamond and Plattner, 2006). For some time, the issue had taken second place to the dominant approach in the study of electoral and party systems, the consequences of electoral rules on the structure of party competition (Duverger, 1954). True, the central question has always evolved around the causal linkage between the two systems, with a clear recognition of a mutual relationship (Benoit, 2001, 2002). Still the prevailing approach in the literature examined the consequences of electoral regulations on party systems, government stability, or executive-legislative relations. The assumption about electoral rules inertia was rooted in the belief that institutions are robust and require external shocks for alteration. The collapse of authoritarian and communist regimes in the third wave and the initiation of electoral law changes in established democracies ushered in a turn to politics in the study of electoral systems: what are the driving forces in the origins and change of election laws? A new orthodoxy posited that the design and reform of electoral institutions is a 1

3 politically endogenous process, whereby political actors have a profound interest in shaping institutions to their advantage. Duverger s approach that electoral formulas structure party systems was turned upside down (Benoit, 2004; Colomer, 2005). More specifically the theoretical claim, supported by empirical evidence, was that fragmented and volatile party systems culminate in preferences for proportional representation (PR), so as to assure political access to the contending parties. In contrast, politics dominated by one or two players tend toward the advocacy and selection of majoritarian rules to amplify their position and restrict access to other political actors (Shugart, 1992; Remmer, 2008). This paper addresses the issue of how the structure of party systems affects the initiation and direction of electoral reforms in the post-communist world. As such, it follows the upsidedown Durvergerian perspective that the configuration of party systems defines the rules of completion. The analysis examines the relationship between the structure of party systems, defined in terms of the effective number of political actors, electoral volatility, and shares of party votes and seats, and electoral reforms, including major and minor types, ranging from the electoral formula to legislative size. The paper consists of three main sections. The first provides an overview of the theoretical perspectives on the linkage between party system and electoral reform. The second looks at the initiation and frequency of reforms in light of the theoretical claim about institutional inertia. The third examines the direction of reform as permissive or restrictive, to evaluate theoretical propositions about party system fragmentation and electoral directionality. Conceptualizing Electoral Reform The collapse of monopolistic power and the ensuing efforts to establish contested politics is fraught with considerable challenges that affect the development of electoral and party 2

4 systems. Emerging democracies give rise to new political forces that lack experience and face an uncertain future, in an inchoate political environment. Reconfiguration of the party systems is a frequent occurrence, as new parties appear, old ones splinter, and voters shift allegiances from election to election. Party systems are fluid, and have been especially so in post-communist states, prone to high rates of electoral volatility and system fragmentation (Mainwaring, 1998; Bielasiak, 2002). In these conditions, the power holders have strong incentives to assure their status as viable political actors. Pressures for alteration in contestation rules appear, to adjust to initial uncertainties and outcomes. There is evidence from various regions of the world that newly competitive regimes turn to reforms of election rules more frequently than the changes in mature democracies (Katz, 2005; Bielasiak, 2006; Reilly, 2007; Remmer, 2008). However, in the face of political fluidity and transition complexity, the reform impulse is subject to two distinct pulls, that of representation and governance. On the one hand, as a response to the chaos of party systems, political entrepreneurs seek to assure success by revising electoral laws towards more relaxed rules that maximize representation, and enhance the prospects of survival in the political game. On the other hand, the fragmented, volatile electoral environment calls for measures that trim party systems to provide better opportunities for stable and effective governance. Political leaders must navigate between these alternatives in their consideration of electoral reform. Three theoretical perspectives are articulated concerning the dilemma of broad representation versus effective governance. Strategic Calculations The first posits a strategic, rational choice motivation for the introduction of electoral change. This view is predicated on the seat and office seeking perspective of power maximization, so that parties undertake instrumental actions to boost their benefits in the 3

5 legislative and policy arenas (Boix, 1999, Benoit, 2004). Electoral reform is a consequence of self-interested politicians responding to party system cues, such as fragmentation and volatility, to cut future losses or expand benefits. In essence, changes in the party systems shape lead to reassessment of electoral rules by politicians committed to safeguard positions of power. High fragmentation and volatility portends the entry of new political players and dispersed political support. In this case, politicians move to protect their own standing by advocating more permissive electoral rules to guarantee access to the legislative arena, i.e. they seek rule changes in the direction of PR away from plurality and majority formulas, or alter PR towards more proportional properties. The tendency towards greater proportionality is accentuated when uncertainty about fortunes in forthcoming elections is especially high (Kaminski, 2002; Pilet, 2007). Risk-aversion explains the predominant global tendency in democratic systems to move from majoritarian to PR systems (Colomer, 2005). Since in emerging competitive polities there is greater likelihood to have unwieldy, fragmented party systems, as in many former communist states, the strategic maximization perspective argues for a predominance of permissive reforms. During the initial period of contested politics, extensive multipartism and high volatility bring forth pressures for more permissive decision rules. Governance Norms An alternative perspective looks not so much to seat maximization as to policy effectiveness as the driving incentive in the initiation of reforms. In this case, parties fortunes depend also on their capacity to implement policies preferred by voters. To stay in power or retain a significant influence in the legislature requires effective policy provisions. Two variants of the policy seeking perspective are evident, the self-interest and the socio-tropic models. 4

6 The first postulates that alternative election mechanisms affect policy outcomes differentially (Benoit, 2002). In PR systems dependent on coalitions and consensus legislation, there is danger of vetoes by smaller parties that undermine the policy agenda. In such cases, policy failures reflect on governing parties, with likely losses at the ballot box. To forestall such outcomes, large parties seek remedy by altering election laws in favor of more constrained systems. In practice, this leads to attempts at more restrictive election rules, even when party systems are highly fragmented (Renwick, 2008). The second variant claims that more general interests also govern the actions of politicians in the process of electoral engineering. The collective interest rather than partisan preferences serve broader ideological perspectives or the norms of democracy, seeking to give reality to the representation of diverse social views (Katz, 1997; Birch et al, 2002). After the collapse of communism, the projection of multiple voices onto the political arena created a spirit of inclusiveness that carried forth in the building of competitive institutions, favoring the adaptation of PR rules. The extensive splintering of the political scene and major shifts in electoral support across post-communist elections presented serious obstacles to stability. High political fragmentation rendered more difficult the formation of stable coalition governments, creating substantial difficulties in agenda setting and policy implementation. Problems of managing society increased public distrust in politicians, parties, and parliaments. The overall effect included punishment at the ballot box by a disenchanted electorate. Since failures in policy lead to poor electoral performance, political entrepreneurs are motivated to gain better control of the fragmented political space through more restrictive electoral procedures. 5

7 Institutional Inertia A third image of electoral development emphasizes inertia, with few changes in election practices. The perspective holds that institutions are robust, and that it is difficult to mobilize support for alterations except in case of external shocks (Nohlen, 1984). Electoral procedures are imbedded in contextual conditions that define choices of electoral systems, whether due to historical experience, social or ethnic requisites, or colonial legacies (Horowitz, 1985; Blais and Massicotte, 1997; Elster et al, 1998). The prospects of reform are further undermined by robust institutional requirements to implement change. Emerging democracies seek to safeguard pluralist procedures of competition by embedding them in foundational documents, making the reform process more tenuous. A variety of provisions are in evidence in the post-communist setting that set a high barrier for reform; anchoring the PR electoral system in the constitution, as in Poland, or subjecting changes to referenda, as in Slovenia or Romania (Millard, 2008; Moniquet, 2008; Fink-Haffner, 2008). In these cases, political entrepreneurs shy away from undertaking tasks necessitating the expansion of political capital in the face of low probabilities of success. So even in conditions of fragmentation and volatility, high levels of uncertainty about reform are likely to produce inaction (Alesina, 2006). The emphasis on maintaining existing contestation rules is likely to be reinforced by evidence of past failures to carry out proposed changes, so the institutional inertia outlook posits a prevailing pattern of fewer reforms than party system fragmentation or electoral volatility would justify. Instead, both the frequency and direction of reforms are likely to be less visible and systematic. While the literature posits a strong relationship between party and electoral systems, it is divided on the impact of the party systems on electoral system changes. The different theoretical 6

8 claims postulate divergent outcomes in conditions of weak party systems, emphasizing power, policy or institutional conditioning that result in either permissive, restrictive, or limited reforms. We turn to an examination of these theoretical propositions through an analysis of how the structure of post-communist party systems impacts on the initiation and direction of electoral reforms. Method: Operationalizing Electoral Reform Case Selection The cases for the study are competitive elections after the founding elections in former communist states in post-1989 East Europe and the post-1991 Soviet and Yugoslav political space. To take into account the quality of democracy question as to the nature of competitive politics in many emerging democratic regimes (Zakaria, 1997; Morlino, 2004; Merkel, 2004), we control for regime type and focus on democratic systems rather than on all former communist states with elections, so as to exclude instances of electorism marred by unfair competition. The central question after all is how unconstrained competition in party systems affects the choices of electoral rules. While there are various methods of operationalizing regime type, we rely on the standard assessments by Polity IV scores. 1 A second issue concerns the time frame for the study. Given the focus on reform, the time horizon covered starts after the founding elections that bring about competitive politics. That is, the first freely contested elections are taken as a base line to gauge subsequent changes in the post-1989 or1991 period for East European and post-soviet states respectively. The result is a data set that includes 74 cases, each case defined as a legislative election for the unilateral or lower chamber, after the first postfounding election and the last election held by the end of December The index enables distinction among three types of political systems, depending on the score in the Polity data set, which classifies polities based on an additive score of authoritarian and democratic features, with -10 to -4 as authoritarian, -3 to 5 as semi-authoritarian, and 6 to 10 as democratic. 7

9 Dependent Variables Another central question concerns the definition of the dependent variable: what constitutes electoral reform? The formal definition of legal changes in the electoral code masks the dilemma of electoral reform significance. The basic question is which changes are sufficiently important to form a new rule configuration that alters the mechanisms of vote to seat conversion and the incentives for political actors (Lijphart, 1994). To assess the differential impact of diverse election rules, efforts have focused on distinguishing between major and minor changes in electoral systems (Lijphart, 1994; Katz, 2005). There is wide consensus that the electoral formula is the critical component in defining the competitive environment (Duverger, 1954; Lijphart, 1994; Katz, 2005, Colomer, 2005). For our purposes, then, we consider changes between majoritarian, mixed, and PR formulas as major reforms that significantly affect the rules of party contestation. A strong case has also been made for the major impact of district magnitude on the vote to seat calculation, postulating a close relationship between electoral and party systems through the medium of district size magnitude. Simply, the higher the number of seats available per district, the better the chance of smaller political formations to win a seat, and thus to participate in political completion (Taagepera and Shugarrt, 1989; Lijphart, 1994; Gallagher, 1991). For that reason, changes in average district magnitude are taken as major alterations that redefine the linkage between electoral and party systems. There are numerous other components to electoral codes that must be taken into account in evaluating the relationship. These minor alterations range across diverse practices (Katz, 2005). We consider the following minor reforms as sufficiently important to affect results: the vote to seats conversion method in PR systems, the number of tiers in the electoral system, assembly size, and legal threshold. More inclusive electoral procedures include two tiers as 8

10 opposed to single tier districting, larger assembly chambers, but lower legal bars for entry into the legislature, as well as shifts in conversion methods from d Hondt and LR-Imeperiali type towards St Lague and Hare calculation methods (Lijphart, 1994, 159). Lijphart s original criterion defined significant modification in magnitude, chamber size, or threshold as a 20% change. This standard is replicated here, so that alongside alteration in electoral formula, changes of 20% or more in the numerical electoral properties take on the characteristics of distinct voting systems. All the changes are then codified as to the direction of the reform as restrictive or permissive. For the purpose of the statistical analysis, the major and minor changes have been combined into a summary indicator denoting the presence of reform and whether it was permissive or restrictive. Direction of Electoral Change The primary questions of the study center first on whether the structure of party systems enhances the initiation of electoral reforms, and second, on the direction of the changes as permissive or restrictive. Different specifications follow from the theoretical claims presented above. Under the strategic self-interest perspective, a large number of political actors increase the prospects for dispersed popular support. In the face of a declining vote share, parties primary interest is to safeguard access to the political system, thereby leading to a rationale advocacy of more relaxed rules (Shugart, 1992; Remmer, 2008). The hypothesis is that high party system fragmentation leads to probability of more permissive rules, while lower fragmentation tends towards the initiation of more restrictive electoral rules. The governance model posits an alternative impulse, based on the premise that fragmentation of the political scene presents greater difficulties for building effective policy coalitions, propelling major political actors to advocate more restrictive rules to eliminate 9

11 smaller political parties. So the hypothesis here is that under high fragmentation, the primary task is to reduce the chaos in the political space by undertaking more restrictive reforms that limit future access to the party system. The institutional inertia model maintains that electoral laws do not change often, for overriding robust institutional barriers necessitates the expenditure of substantial capital in the face of risky outcomes. Moreover, prolonged duration of existing rules acquires a focal point that is difficult to overcome since it involves greater political costs (Alesina, 2006). In this case, the hypothesis holds that despite fragmentation, the prevailing pattern makes reform less likely over subsequent electoral cycles. The probability of changing the rules is also affected by electoral stability or volatility (Remmer, 2008, 10). Consistency in the outcomes across elections reinforces institutional inertia, as the winners have no motives for altering the rule of the game. The expectation is that low volatility will signify no or few changes in the election codes. Instances of high volatility, in contrast, render all political actors subject to greater uncertainty and present incentives to minimize the risks by introducing reforms. The direction of the changes is bounded by the position and preferences of political actors, so that strategic calculations of self-interest dictate preferences for permissive change, while concern with governance capability redirect preferences towards more restrictive elements in the electoral law. The hypothesis is that volatility encourages reform, while stability reinforces status quo in the electoral rules. Independent Variables The independent variables concern the structure of party systems, as the effective number of political actors, electoral volatility, and shares of party votes and seats. The configuration of party systems is determined in large part by the number of functioning political actors, for it 10

12 structures the placement of parties along the political space, and in turn forms choices for the electorate. In that sense, the number of effective parties is a good indicator of political fragmentation. We employ the standard distinction between electoral (ENPV) and parliamentary (ENPS) party systems, to determine the effect of vote and seat fragmentation on the initiation and direction of reforms. 2 Two additional indicators are employed in the analysis as evidence of power concentration in the party system, namely the percent share for largest party in terms of vote and seats, as well as the percent of seats controlled by the two largest political parties in the unicameral or lower house of parliament, as a means to gauge the capacity of dominant political players to revise the electoral systems. Finally, we use the Pederson index of electoral volatility (EVV), which concentrates on changes in the voting share for parties in consecutive elections. 3 The premise is that consistent voting results over time as opposed to significant swings in the electorate s preferences form stable or inchoate party systems that affect incentives for reform. To reflect the belief in the literature that electoral changes may be driven by economic factors or the duration of the democratic regime, we turn to two principal control variables. The first concerns an evaluation of economic performance through inclusion of GDP growth rates, lagged for years prior to the election year (World Bank Development Indicators). The second is a measure of electoral cycles, defined by the electoral sequence since the transition from communist to democratic regime. 2 The measure of effective number of parties uses the Laakso and Taagepera formula: N = 1/ Σ p i 2 where N is the effective number of parties and p i is the fractional share of votes or seats for the i-th party. 3 Electoral volatility is calculated as V = ½ v p, t v p,t-1 where v p, t stands for the percentage of the vote obtained by a party at election t, and v p,t-1 for the percentage in the previous election (Pedersen, 1979). 11

13 Party Systems and Electoral Reform: Results We begin with an assessment of electoral reforms in light of the third model s assertion that institutions are robust and difficult to revise even in conditions of extensive fragmentation and volatility. We examine the frequency and temporal pattern of the changes, the type of reforms, and the direction of alterations through bivariate descriptive statistics, as well as statistical models to take a closer look at the causes of reform. Initiation and Frequency of Reform An overview of the changes in election rules is the post-communist world is presented in Table 1 for competitive elections during Reforms were introduced in 18 countries, and applied to 33 different election cycles, out of a total of 74 elections in the sample. The only country in the region without electoral tampering was Estonia, whose unusually complex code has remained in place since its initiation for the 1992 election. Even in this case, if the founding election is moved back to 1990 when Estonia conducted a free electoral contest while still part of the Soviet Union, there is a change in formula between the 1990 STV rules and the three tiered, linked PR system employed since Table 1 about here The changes in the other post-communist states ranged across all properties of the electoral codes, from alterations in the electoral formula to alterations in legislative size and thresholds. Many of these episodes involved reforms in multiple features of the electoral code, within the same election cycle, with a total of 52 voting rule modifications clustered in 33 election cycles. While these changes are examined separately in our initial analysis, in the statistical models we treat each election cycle as a unit, avoiding duplicate counts on the dependent variable. 12

14 The majority of changes (22) apply to one specific feature of the electoral law, concentrated in reforms of electoral formula and the legal threshold. Both of these measures have the advantage of transparency, providing a clear signal as to the intent of the reform and offering inducements to political actors interested in change. In addition, there are 11 episodes of more complex reform clusters that affect several different properties at the same time, even while discounting changes due to the natural effects of alterations in formula type. When there are multiple reforms in one election cycle, they are generally unidirectional, moving to either restrictive or permissive rules. The former tend to be concentrated in the early period, after the initial burst of political actors onto the post-communist political space leads to constraint efforts (e.g. Poland in 1993), while the latter are often associated with subsequent second transition openings of the political system (e.g. Croatia in 2000). In addition, in several instances, the nature of reform is not as transparent, involving changes in the same reform package that are both permissive and restrictive; these have been coded as to the net effect of the reform for each election cycle (e.g. Slovakia 1998 and Poland 2001). We return to the directionality issue below, at this point we emphasize the relatively high frequency of reforms after the founding elections of post-communism. The prevalence of electoral innovations in the region is confirmed through comparison with changes in other parts of the world. In regard to the major reform of electoral formula, 14 such episodes have taken place in the established democracies between 1950 and 2002, in both legislative and executive arenas (Katz, 2005); 6 legislative reforms in mature democracies between 1959 and 2006 (Renwick, 2008); and 6 in the Asia-Pacific region between 1990 and 2004 (Reilley, 2007). In the post-communist countries in , there were 10 outright formula changes and three additional changes in the single member segment of mixed electoral systems. Expanding the 13

15 comparison to all types of reform, Lijphart (1994) identified 30 for in 27 established democracies, in contrast to the 33 reforms instituted in 22 post-communist states during The comparative evidence, then, demonstrates the extensive nature of reforms, and casts a shadow on the claim that institutional inertia is a dominant aspect of political development even in condition of inchoate, volatile political systems. In emerging democracies, when party systems are unsettled and governmental effectiveness is in doubt, there are powerful inducements for the provision on new electoral rules to alter political conditions. Though reform prospects were reduced by imbedding electoral provision in legal requirements of constitutional amendments, public referenda, or supermajority legislative votes, reforms have been a major feature of the post-communist landscape. There were ten reforms of the electoral formula in six states, but concentrated in three countries. Mongolia is responsible for three shifts, with two consecutive reforms in which the latter reversed the first, from single member run-off to multimember plurality, and back to SMM. The other cases of multiple formula alterations took place in Macedonia and Ukraine, but these were unidirectional, from SMD to Mixed to PR systems. In addition there were three formula adjustments within the single member component of mixed systems, but again one set involved a turnaround: Lithuania first moved from majority run-off to plurality rules, than reverted back in the following election. While this concentration may produce the impression that changes in formula were not as prevalent as initially noted, the number of affected countries is still twice that of the alterations in established democracies. If we embrace a more expansive definition of major reforms to include district magnitude, then the number of adjustments in elections law increase significantly, as there were eight such changes due to redistricting and eight additional ones due to reform in other electoral 14

16 properties, e.g. size of the legislative chamber. These important revisions in the conduct of elections were supplemented by additional alterations in other minor reforms of the electoral code. Here the most pervasive instrument is the application of the threshold, with 19 revisions, by far the most common element of electoral engineering in the post-communist environment. Again, this is tied to the transparency of the change, which is much easier to interpret than other modifications in the electoral system. For that reason, increases in the legal bar were an especially popular mechanism in East Europe and the former Soviet Union in the initial phase of democratization, when political chaos characterized the newly opened political space, and provided a catalyst to restrain the political noise through more restrictive rules. Table 2 about here The inchoate nature of the new politics in the aftermath of the 1989 breakthrough is a factor in the temporal trend of reforms. There is a high concentration of changes in the first two post-founding election cycles, with a probability of reform at 77% in the second wave of elections, 50% in the third cycle, and a drop off to slightly above 30% in the subsequent elections. The propensity to initiate election reform is skewed towards the first voting contests, with 18 out of the 33 reform clusters during the first two cycles, and the other 13 reforms spread across the subsequent electoral periods (Table 2). This trend is evident also for the major reform types, with eight alterations in electoral formula at the forefront of competitive politics, with the remaining five in the subsequent three elections; a pattern repeated for the changes in district magnitude. The chronological path of reform implementation mirrors the extent of party system fragmentation, which is higher during the years immediately following the fall of the ancient regimes and levels off during the last three election cycles, although the level of volatility remains high for much of the period. So at least in terms of comparative trends, countries that 15

17 undertook reform had higher average volatility and fragmentation in the preceding election cycle than those that did not undertake reform. However, a multivariate probit analysis for the entire post communist period does not find any meaningful effect of party system shape on the propensity to adopt reforms. Table 3 about here This holds for the lagged measure of party fragmentation by votes and seats, and for control of parliamentary majorities. In both the votes and the seats models, the estimated effect of fragmentation measured by the effective number of parties shows no substantive or statistically significant effect. Electoral prospects are more likely to influence the likelihood of reforms. The effect of electoral volatility is large in the first few election cycles and statistically significant to the.01 level. In the second election cycle, the first in which we can observe change, the predicted probability of reform for countries with average levels of volatility is.63 (with all other independent and control variables held at their mean or modal values). Those with levels of volatility one standard deviation below and above the mean showed predicted probabilities of.17 and.95 respectively. This very large substantive effect all but disappears in the fourth election cycle and beyond. In the model based on votes, the vote share of the largest party has a positive and statistically significant effect, which fades in later election cycles. Larger vote shares are associated with increased likelihoods of reform, all else being equal. In the second election cycle, again the first in which we can observe reform, the predicted probability of reform for countries with average vote shares for the largest party is.66 (with all other independent and control variables held at their mean or modal values). Increasing or decreasing the largest party s vote share by one standard deviation yields predicted probabilities of.90 and.33 16

18 respectively. The effect over time is more robust than for electoral volatility but of somewhat smaller magnitude. Voting trends rather than party system fragmentation are more likely to affect the chances of reform. Direction of Reforms Nonetheless, reforms of electoral systems have been a consistent feature of the democratizing experience in the post-communist world. The central question is whether the initiation of reform favors permissive or restrictive innovations. The empirical evidence on the direction of reform is mixed (Table 4). On the most inclusive level, for all major and minor reforms, there is no overall pattern favoring one type of reform over the other; in fact there is an equal division with 17 restrictive and 16 permissive electoral innovations. Given the absence of a unidirectional, systematic trend, we cannot attribute the implementation of reforms to either the strategic seat maximization or the governance effectiveness motivations. Rather both models seem applicable at different times, for different reforms. As a chronological trend, there is some evidence that restrictive laws were concentrated in the first two post-founding voting cycles, as a response to the inchoate nature of the political system at the start of democratization, indicating a desire to impose greater efficiency. At the same time, however, there is an evident presence of permissive reforms throughout the period, with a flurry of activity during the third cycle, indicating concern with strategic calculations of representation. Table 4 about here There is some indication from the bivariate pattern that the shape of party systems does affect the direction of change. Systems that introduce more restrictive electoral rules have higher levels of party system fragmentation in the prior election cycle in terms of votes (6.1 average) and seats (4.4), and an increase in fragmentation over time. In comparison, reforms in a 17

19 permissive direction occur in conditions with fewer effective numbers of electoral (4.4 average) and parliamentary parties (3.0 average), and are undergoing a decline in the former measure. From this evidence, we can deduce that the high fragmentation of post-communist parties tends to favor measures that seek to introduce a more manageable political environment, giving greater credence to the effective governance perspective. However, permissive reforms are associated with higher levels of electoral volatility (27.2 average) than restrictive reforms (23.6). In both trends, volatility is still quite significant, so that that strategic position that high volatility brings about more permissive rules, and the governance perspective that high volatility results in more restrictive electoral applications are possible. In order to clarify the timing and direction of electoral innovations, we turn to a multinomial logit model, which allows for a joint estimation of the likelihood of both direction of reform. We divide instances of reform into restrictive and permissive categories in order to investigate the causal factors associated with each type of reform. Table 5 shows two multinomial logit models, one based on measures derived from votes and one based on measures derived from seats. The baseline outcome is no reform, so the models for restrictive and permissive reform are in relation to no reform. Just as in the probit models, we have included a continuous variable for election cycle and interaction terms with the variables of theoretical interest. Table 5 about here The results for electoral volatility by votes (EVV) are very similar to those in the probit model. Interestingly, the effect is almost identical for both restrictive and permissive reform, meaning that volatility provides no leverage for ascertaining the likelihood of restrictive over permissive electoral changes. Rather higher levels of volatility in the previous electoral cycle are 18

20 associated with a greater probability of both types of rule changes, towards more open and more closed systems. The coefficient is statistically significant for the restrictive type, and while the coefficient for volatility under permissive reform is not significant at the.05 level, it only just falls short of that critical value. As a specific example of the relationship, Figure 1 shows the predicted probabilities for each value of volatility in the votes model for the second election cycle; all other variables are held at their mean or modal values. The higher likelihood of permissive reform reflects the greater propensity toward that type of reform in the first election cycles. 4 Most of the cases fall between 15 and 35 on volatility, and the mean is 25. Figure 1 shows the strong impact of volatility on the likelihood of both types of rule modification. Figure 1 about here These results support aspects of both the strategic perspective s contention that high volatility increases the likelihood of permissive reform and the governance perspective s claim that high volatility brings about restrictive reform. Additionally, the negative sign on the interaction term suggests increasing inertia in later election cycles. In relation to the other variables included in the model, volatility is the strongest predictor of reform, but does not provide solid information on the direction of the change as either permissive or restrictive. In the votes model, the effective number of parties produces results in opposite directions, but the standard errors are very large, throwing doubt on any substantive interpretation. However, in the seats model, there is a strong negative impact on the likelihood of alterations towards more permissive election rules. Higher numbers of parties are associated with a lower likelihood of permissive reform, although the strength of the effect decreases in 4 This may be, in part, an artifact of the modeling choices regarding election cycle discussed above. The reforms are split evenly between restrictive and permissive reform in the second election cycle, but permissive reforms predominate in the third electoral cycle. The choice of a continuous specification of electoral cycle instead of a set of dummy variables forces a more linear relationship between reform and election cycle. 19

21 later election cycles, and the effect is gone by the fourth cycle. Therefore, these results do not provide support for claims that party system fragmentation affects the likelihood of permissive or restrictive reform. In both the votes and the seats model, the effect of the largest party vote share is small and not statistically significant, however, as in the probit seats model, the multinomial logit seats model slows a statistically significant, negative relationship between the vote share of the largest two parties and the likelihood of reform. In the multinomial logit seats model, the effect is only statistically significant for permissive reforms. Greater vote shares for the two largest parties are associated with a decreased likelihood of permissive reforms, although the size of the impact decreases in later election cycles. Interpreting the impact of the election cycle variable is made difficult by the interaction terms with the volatility and fragmentation measures. Even though the coefficients for election cycle in the votes model are positive, the combined net effect of election cycle and the interaction terms including the cycle measure is negative when all variables are held at their mean. The net effect of election cycle is shown in Figure 2, which depicts the predicted probability for each type of reform by successive elections. Consistent with the inertia perspective, the probabilities of both types of reform decrease dramatically over time. Figure 2 about here Types of Reform: Frequency and Direction A more evident temporal pattern as to the introduction of permissive or restrictive criteria emerges in conjunction with specific types of reform. For the most significant reform, that of formula, the direction is markedly more towards the permissive side, with nine alterations as opposed to four in the opposite direction. On this dimension, the evidence points to more 20

22 frequent consideration of strategic calculations for assuring representation through more relaxed rules. Most of the changes in the formula mirror the long-term global path towards PR defined by Colomer (2005). A closer examination of the evidence outlined in Table 1 reveals that the trend is even more pronounced when discounting reversals in subsequent elections. In that case, the majority of formula reforms have moved towards proportional representation, whether in Bulgaria at the start of the post-communist transition or later on in the Ukraine, Macedonia or Croatia in conjunction with pressures associated with the second transition from dominant party systems. In all these instances, the reform has moved away from single member plurality or majoritarian systems to mixed or PR formulas, and then from mixed to full PR. Two countries, as noted previously, have initiated formula reforms only to reinstitute the previous method in the subsequent election. Both in Lithuania and Mongolia, the final disposition in the experimentation with formula alterations was towards a more permissive electoral system. In fact, of the eight countries applying revisions to their formula, only Albania 2005 application of plurality to its SMD segment and Romania 2008 introduction of a MMP have culminated in more restrictive election practices. These last two more recent changes reflect other attempts in the region to bring about more restrictive systems, as for example in the arguments advanced in Poland for the replacement of PR with mixed or majoritarian systems (Millard, 2008). In these instances, the justification for moving away from pure PR is to facilitate the work of the legislature, overcome weak government, strengthen the main parties and consolidate the party system. The effort, successful or not, is clearly a reflection of the dominant parties concern with streaming the unruly competitive space and facilitate more efficient government. 21

23 Similarly, the governance model s assertion that effectiveness is a critical component in the selection of electoral reforms is borne out by long-standing practices concerning the electoral threshold and the vote to seat conversion in proportional representation systems. Revisions in these two properties of election codes obviously aim at restrictive practices that signal a desire to trim the party systems. By far the most common reform in that regard is the introduction or increase in the legal bar for entry into parliament, affecting requirements for single parties and for pre-election coalitions. Once again, the practice was especially prevalent during the first two post-founding elections cycles, when the explosion of new political actors onto the democratic competitive field rendered the legislative agenda problematic. The paradigmatic case here is Poland between 1991 and 1993, when the appearance of dozen of parties in the electoral and parliamentary space made stable governance impossible, leading to the imposition of a 5% threshold for parties and 8% for coalition at the lower tier, and an increase from 5 to 7% at the upper tier of the unlinked PR system; accompanied at the same time by a shift to the more restrictive d Hondt conversion method. Similar practices took place throughout the region in the first part of the 1990s, with higher threshold requirements imposed in the Czech Republic, Hungary, Latvia, Lithuania, Romania, and Slovakia (Table 1). The only significant exception to the reductive effect of threshold reforms occur in Croatia 2000 and Georgia 2008, when the appearance of new political forces in the aftermath of the erosion of strong-man rule provides a strong impetus to open up the political process and assure representation. The greater propensity to institute more restrictive demands through the legal requirement and the vote to seat conversion rules is countered by the more permissive direction of changes in district magnitude. On this dimension, taking into account district, assembly size, and tier changes that affect the magnitude of the electoral system, there were 12 permissive 22

24 alterations in contrast to just four restrictive ones. Even when we discount the changes in magnitude due to the shifts from SMD or mixed to PR systems, there are still significant shifts in magnitude within PR, the most evident in Slovakia s adaptation of a nationwide district of 150 in 1998 and Montenegro s return to a single district of 30 in Conclusion In view of these changes, there is an obvious discrepancy, or at least a duality, in the direction of reforms in the post-1989 period. The post-communist states have embraced both permissive and restrictive strategies of electoral system transformation. On the whole, major reforms in formula and district magnitude have moved the election practices towards more permissive rules. These steps, however, are countered by other, significant restrictive innovations, most evident in threshold and conversion rules. For the democratic cases of postcommunism, then, some states have initiated reforms that are permissive and suggest predominant concerns with strategic calculations. Other countries have favored more restrictive adjustments in electoral properties, pointing to a greater concern with policy efficiency. At first glance, the trends do not produce clear evidence how the shape of the party system informs the initiation and direction of electoral reforms. In view of both permissive and restrictive electoral changes present in the post-communist states, the question as to what determines the decision to reform and in what direction remains uncertain. Still, throughout much of the region, there is a propensity towards electoral law fetishism, evident in the sequence of reforms, as well as the continuing attempts at reform. Modifications of electoral codes are embraced as remedies for the ills of the party systems, either to assure representation in an unstable political context or to provide better efficiency in governance. But the problem may well be deeper that election decision rules; associated rather 23

25 with the extent and complexity of the post-communist transition that generates an inchoate party system populated by weak political parties. The solution to the dilemma may well rest beyond the alteration of electoral systems, through the gradual emergence of winners and losers in the competitive process bound by established rules, rather than the manipulation of electoral institutions. 24

26 Table 1. Electoral Reforms in Unicameral or Lower Legislative Chamber Country Year Electoral Reforms Reform Direction Albania 1997 Decrease in threshold From majority run-off to plurality in SMD segment - Bosnia 2000 District change: decrease in average district magnitude - Introduction of threshold - Single to two tiers PR Introduction of threshold in upper tier - Bulgaria 1991 From Mixed to PR system + Reduced number of deputies - Croatia 2000 From Mixed to PR system + Increased number of deputies + Decrease in threshold + Czech R District change: increase in average district magnitude + Increased number of deputies + Introduce threshold for coalitions District change: decrease in average district magnitude - Increased threshold for coalitions - Two tiers to single tier PR - Georgia 2008 District change: increase in average district magnitude + Decrease in threshold + Reduced number of deputies - Hungary 1994 Increase in threshold - Latvia 1995 Increase in threshold - Lithuania 1996 Increase in threshold From majority run-off to plurality in SMD segment From plurality to majority run-off in SMD segment + Macedonia 1998 From SMD to Mixed system From Mixed to PR system + Moldova 2001 Increase in threshold - 25

27 Table 1. Electoral Reforms in Unicameral or Lower Legislative Chamber/continued Country Year Electoral Reforms Reform Direction Mongolia 1992 SMD to multi member plurality MMP to single member plurality SMD to multi member plurality + Poland 1993 District change: decrease in average district magnitude - Change in seat allocation formula - Increase in threshold District change: increase in average district magnitude + Change from two tiers to one tier PR - Change in seat allocation formula Change in seat allocation formula - Romania 1996 Introduce higher threshold for coalitions Increase in threshold PR to multi member proportional - Slovakia 1992 Change in lower level seat allocation formula - Increase in threshold District change: increase in average district magnitude + Change from two tiers to one tier PR - Increase in threshold for coalition Decrease in threshold for coalition + Slovenia 2000 Change in seat allocation formula - Ukraine 1998 SMD to Mixed electoral system Mixed to PR electoral system + Decrease in threshold in Mix PR to full PR - Yugoslavia Montenegro 2000 District change: increase in average district magnitude + 26