Party competition and electoral reforms: why do governments initiate a reform?

Abstract Most accounts of electoral reforms focus on successfully implemented reforms to explain how electoral context shapes the incentives of political parties, paying scant attention to the cases where governments fail to implement their preferred system. This article takes a step back in the electoral reform process and examines when and why governing parties initiate electoral reforms. In doing so, it focuses on how the electoral context can affect the electoral bases of the incumbents and their main competitor. This novel account expects that governments initiate electoral reforms depending on whether small or new parties draw votes from their own vote base or from that of their main competitor. Using an original dataset of electoral reform attempts from 32 parliamentary democracies between 1945 and 2015, this article shows that ruling parties are more likely to initiate a restrictive reform when small parties draw votes from their electoral base, but a permissive one when small parties draw more votes from their main competitor.

districts, making the electoral system more permissive, but failed to garner the required two-thirds majority support in the parliament (Fink-Hafner and Krasovec 2014). 1 These outcomes constitute an important part of the electoral reform process and are essential to any endeavour to understand why politicians decide to revise the electoral rules.Their exclusion risks introducing bias to the empirical analyses, since a 'no reform' category does not necessarily mean no reform was attempted.This article takes a step back in the electoral reform process and asks when and why governing parties initiate permissive or restrictive reforms.
In order to answer this question, I develop a strategic model to specify the conditions that change the incentives of incumbent parties regarding the electoral system. 2 I build on the dynamic models of electoral reforms which focus on the relationship between party system structure and electoral reform types (Bielasiak and Hulsey 2013;Colomer 2005;Harfst 2013;Nunez et al. 2017;Remmer 2008;Shugart 1992).This emerging literature claims that electoral volatility and party system fragmentation give rise to permissive or restrictive reforms.Two sets of explanations have been advanced to explain the relationship between party system instability and electoral reforms.The first one argues that electoral reforms reinforce the existing trends in the party system: permissive reforms are likely when party system fragmentation is high, i.e. the number of relevant parties is high in an electoral cycle, and restrictive reforms are likely when it is low (Colomer 2004;Colomer 2005;Remmer 2008).The second posits that electoral reforms reverse the existing dynamics in the party system: restrictive rules are introduced when electoral volatility and party system fragmentation are high (Bielasiak and Hulsey 2013;Nunez et al. 2017;Riera 2013).These rival explanations find support for restrictive reforms in their analyses, however, neither provides a compelling motivation for permissive reforms that have an empirical support.
This article argues that the relationship between party system instability and electoral system changes is more nuanced, and the direction of change depends on how party system instability affects the governing party's vote base.The existing work does not consider that party system instability can affect the leading parties in the government and opposition asymmetrically; in other words, that small and/or new parties can draw votes from either governing party's vote base or that of its main competitor.Considering this asymmetrical impact, I argue and empirically demonstrate that the political party in power would prefer a restrictive reform when the number of small or new parties drawing votes from its vote base is high, to induce voters to vote strategically for the viable larger party.In contrast, the governing party would prefer a permissive reform, which opens the system to small and new parties, when there are more parties to draw votes from its main competitor, to weaken its vote base.
In order to empirically test the nuanced relationship, this study uses an original dataset of party competition and electoral reform attempts.The dataset covers all reform attempts affecting proportionality of electoral systems in parliamentary democracies in the period between 1945 and 2015, diverging from the existing studies which include only successfully implemented reforms.To calculate the number of parties drawing votes from the leading parties in government and in opposition, I offer a new method of counting relevant parties: fragmentation in the ideological clusters.Assuming that small parties can draw votes from the larger party, which is ideologically proximate, I calculate the effective number of parties which are ideologically closer to the leading party in government and in opposition separately for each electoral cycle.This variable provides a better and politically relevant measure to understand the changes in potential electoral bases of the main parties and the strategic calculations of party in power.
In the empirical analyses, I first tested hypotheses of two main explanations in the literature with the reform attempts dataset but did not find strong support for them.Multinomial logit estimations suggest that electoral volatility increases the propensity for electoral reform attempts, but it does not predict their direction.Also, overall party system fragmentation does not have a relationship with electoral reform attempts as expected by the two main explanations.However, the fragmentation in a government's ideological cluster, consistent with my argument, does increase the likelihood of restrictive reform attempts while decreasing the probability of permissive ones.I also find support, albeit weak, for my argument that permissive changes are initiated to weaken the opposition: they are more likely when the incumbent can increase the fragmentation in the opposition's cluster more.These findings point to the need to dig deeper into the electoral reform process and contextualise the relationship between party system and electoral reforms.

Previous work on electoral reforms
The conventional wisdom about electoral systems has been that they remain stable once put into use (Benoit 2004(Benoit , 2007)).This view is reflected in the early studies of electoral reforms which focussed on the rule changes during the democratisation process.The underlying idea is that electoral systems do not change unless there is an exogenous shock altering the preferences of political actors (Boix 1999;Calvo 2009;Birch et al. 2002).In the adoption of proportional representation (PR) in Western Europe, for example, the exogenous shock was the suffrage expansion.The established parties used PR reforms to contain the working-class mobilisation and the rise of socialist parties in response to the (demands of) universal suffrage (Ahmed 2013, Boix 1999, Pilon 2013).Boix (1999) focussed on the electoral threat to argue that PR was adopted by countries where right-wing parties were fragmented, and the socialist party was strong; whilst others focussed on the existential threat that working class and socialist parties posed to the institutions of liberal democracy and capitalism (Ahmed 2013).The latter studies argued that ideological radicalism of socialist parties posed more of a threat to the established political parties, leading to the PR reforms (Ahmed 2013;Pilon 2013).Given the exogenous shock, these studies did not need to theorise about the timing of PR reforms.
The perception of stability in the electoral system has also been reflected in the dominant methodological approach to study electoral reforms.As Leyenaar and Hazan (2011) identify, the general methodological trend has been single case studies of reform episodes.The timing of electoral reforms is less pronounced in the case studies: as they do not address the question of when electoral reforms occur, but instead examine the goals of political actors and their reform choices during an ongoing reform debate (Santucci 2018).Most case studies examine the strategic calculations of reformers regarding electoral rules (Benoit and Hayden 2004;Bawn 1993).Some others argue that the flaws of the electoral system make it susceptible to electoral reforms (Shugart 2001;Shugart and Wattenberg 2001) but leave the timing of reforms to contingent events such as corruption scandals, rather than systematically studying the contextual factors.
The wave of electoral reform attempts over the last two decades has changed this trend by showing various ways in which electoral systems are designed and altered.Studies have also found that seemingly minor changes in the electoral threshold, district magnitude, and assembly size do have significant impact on the seat allocation to political parties (Carey and Hix 2011;Lijphart 1994).These developments rendered the electoral reform definition as the wholesale replacement of an electoral system too narrow (Jacobs and Leyenaar 2011;Nunez and Jacobs 2016).With a broader definition, scholars studied how contextual factors, such as party system size, can affect the stability and change in electoral rules in broader samples of countries over time.
It is possible to group the explanations in the current literature along Shugart's (1992: 209) two rival hypotheses about the relationship between party system size and electoral reforms.The first group focuses on his hypothesis that permissive electoral reforms are adopted when party system fragmentation is high, whereas restrictive ones are preferred when the fragmentation is low (Colomer 2004 and2005;Remmer 2008;Shugart 1992).In other words, the first group argues that electoral system reforms reinforce the existing trends in party systems (Remmer 2008).These studies view party system fragmentation as a source of uncertainty and an electoral risk, since there are more parties competing in elections, making existing parties unsure about their electoral strength.With increased competition, parties are motivated to avoid major seat losses and have an incentive to replace the existing electoral system with a less risky alternative: a permissive reform (Colomer 2004;Remmer 2008).The rationale is that permissive rules produce multiple winners and decrease the chances of a devastating election loss.In other words, adopting a permissive reform is a defensive strategy to prevent electoral losses (Santucci 2018).In contrast, when the level of fragmentation is low, parties adopt restrictive reforms to consolidate their gains from a low level of competition (Remmer 2008;Shugart 1992).
The second group argues that electoral reforms reverse the trends in the party systems (Bielasiak and Hulsey 2013;Harfst 2013;Nunez et al. 2017;Riera 2013;Shugart 1992).In other words, restrictive reforms are likely when party system fragmentation is high, whereas permissive reforms are more likely when party system fragmentation is low (Shugart 1992).Scholars in this group paid more attention to the relationship between restrictive reforms and party system fragmentation.They argue that highly fragmented party systems are associated with governability problems; as it becomes difficult to form durable governments, pass laws and make policies when small parties have veto power in the parliament.A restrictive reform is adopted to remedy the governability problems and to reverse the trend of increasing numbers of parties (Bielasiak and Hulsey 2013;Harfst 2013;Riera 2013).In contrast, permissive reforms are more likely when party system fragmentation is low since it is safe to open up the electoral competition when there are few parties (Shugart 1992: 209).
Within this school, electoral volatility is seen as incentivizing politicians to respond to the shift in voter allegiance, the emergence of new parties, and the disappearance of existing ones, by altering institutions to ensure their viability.While many cross-national studies find a positive relationship between electoral volatility and reforms, they do not offer any insight into the type of electoral reform adopted (Bielasiak and Hulsey 2013;Nunez and Jacobs 2016;Remmer 2008;Riera 2013).Nunez et al. (2017), on the other hand, argue that the emergence of new parties challenges the advantaged position of existing ones and creates an incentive to prevent new competitors from entering the parliament.Therefore, new party volatility, resulting from the vote shifts towards the new parties, makes restrictive reforms more likely because existing parties can act like a cartel and adopt restrictive reforms to exclude the new ones (Nunez et al. 2017 p. 383).
The empirical findings are inconclusive in terms of favouring one explanation over the others; there exists only partial support for them.While Remmer (2008) finds that restrictive reforms have a negative relationship with party system fragmentation in Latin American cases, suggesting winners want to consolidate their gains from low levels of competition.Harfst (2013) finds a positive relationship with party system fragmentation and the success of restrictive reforms in Central and Eastern Europe, suggesting that electoral reforms reverse the party system fragmentation.Similarly, Nunez et al. (2017) find that restrictive reforms are used to reverse electoral volatility stemming from new parties in European cases.Neither explanation provides a compelling motivation for permissive reforms that is empirically supported.
One of the reasons for the mixed results might be the inattention to the initial electoral institutions.Blais et al. (2005) highlight the importance of existing electoral institutions in affecting the electoral reform decision, since the uncertainty faced by both voters and parties is different depending on the initial electoral rule. 3The variation in the initial electoral systems can affect both the strategic calculations of parties and the alternative directions that electoral systems can take.For instance, electoral competition in a multiparty context can be risky especially if the third party is strong within a plurality/majoritarian system.But PR systems produce partial winners and therefore vote swings and emergence of small or new parties do not pose a comparable risk and uncertainty.Moreover, the pre-existing electoral institutions constrain the possible tools that can be used to increase or decrease the proportionality of the system.While in PR systems, politicians can increase the electoral threshold, switch from electoral formulas based on quota to those based on divisors, introduce additional tiers benefiting larger parties in addition to changing the district magnitude, to make the system more restrictive, in majoritarian/ plurality systems, such tools are quite limited. 4In fact, there are studies focussing solely on countries using plurality systems to explain when they are replaced (Blais 2008;Shugart 2008).To ensure institutional comparability, I exclude the cases using majoritarian rules in this article.
Second, as Nunez and Jacobs (2016) underline, most existing statistical analyses focus on factors that facilitate electoral reform but there are many reform debates that result in failure.These failed reforms are, indeed, an important part of the process (Bol 2016;Levick 2016;Rahat and Hazan 2011).Their omission from the empirical analyses risks missing a substantial part in which politicians had incentives to change the electoral system.For a complete analysis of the party preferences, the inclusion of unsuccessful reforms is needed so that a relevant comparative base in the empirical analyses is achieved: cases where government initiated an electoral reform and where they did not.
Finally, a further examination of the relationship between the party systems and electoral reform is needed.Existing explanations assume larger parties are in a position to change the electoral system in parliament, yet they fail to consider scenarios where party system fragmentation affects larger parties differently and where their gains from electoral reforms are not aligned.Given that electoral institutions are redistributive (Tsebelis 1990), these scenarios are highly likely.In particular, if we consider that larger parties are the main contenders in elections which compete against each other to form the government, the asymmetrical effects of party system instability become more pertinent to understanding electoral reforms.Indeed, case studies have long highlighted the different electoral institutions preferred by the largest parties (Bawn 1993;Tsebelis 1990).Since electoral rules are modified in a top-down manner, and it is usually the governing majority which modifies the electoral rules (North 1990;Pilet 2007), I focus on the governing party and its incentives to change the electoral system.

Party competition and electoral reform attempts
While electoral competition is higher in fragmented party systems due to new and small parties, it does not call for a uniform institutional response in the form of restrictive or permissive reforms.New and/or small parties can gain their votes at the expense of one of the leading parties, weakening the vote base of that party.This differential impact of fragmentation, in turn, affects the competition between the leading parties and shapes the preferences of the incumbent accordingly.Therefore, central to my explanation is the asymmetric impact of party system fragmentation.I assume that parties want to maximise their chances of forming (or taking part in) a government and their power within a possible governing coalition.Their preference over alternative electoral rules, therefore, reflects this goal. 5It follows from this assumption that the incumbent party faces two sources of competition in an election: the competition with the largest opposition party over the government formation; and the competition with the small parties over the power within the governing coalition. 6Party system fragmentation can affect both competitions faced by the incumbent.
However, the party system fragmentation, on its own, does not provide any cues about how the competition for government formation between the largest two parties is affected by other, possibly smaller, parties.High levels of party system fragmentation signal an increasing number and weight of small parties, which may stem from vote transfers from either one of the leading parties to smaller ones.For the incumbent party to evaluate the alternative systems, it is important to capture at whose expense the small parties gain votes.The ideological proximity of small parties to either leading party can signal future vote shifts to small parties if the status quo rules are maintained.In this respect, the placement of small and new parties on the ideological spectrum, and their distance from the main government and opposition parties are essential to the incumbent's decision over electoral reform.
For instance, fragmentation close to the governing party's ideology signals a high degree of competition in this ideological cluster.With the status quo electoral rules, this type of competition is likely to decrease the seat potential of the incumbent party as well as increasing the likelihood of future vote shifts to the small parties located close to itself.In such a context, the incumbent party can have an incentive to respond to these shifts with institutional solutions, such as initiating a restrictive reform which, in general, encourages voters to coordinate on the viable political parties close to their ideological preferences.
In this scenario, the incumbent is not likely to initiate a permissive reform since permissive rules would help small parties to enter the parliament and consequently reduce the legislative power of the governing party.In addition to these short-term effects, permissive electoral reforms may also lead to further fragmentation within ideological blocs by encouraging party splits or emergence of new parties.This would ultimately mean an increase in the number of parties close to the governing party, and hence it would have no incentive to facilitate it.The following hypothesis addresses the relationship between the degree of competition in the governing party's ideological cluster and its incentives to initiate reforms: Hypothesis 1: The probability of restrictive reform attempts increases with the effective number of parties in the ruling party's ideological cluster, while the probability of permissive reform attempts decreases with it.
As highlighted above, permissive reforms are likely to increase the number of small parties and weaken the chances of forming a cohesive government.This type of reform, therefore, can serve to weaken the opposition, the incumbent can use it to prevent the consolidation of the opposition by increasing the degree of competition in the opposition party's ideological cluster.Renwick et al. (2009) show that one of the aims of the centre-right coalition during the 2005 reform in Italy was to make the centre-left coalition more dependent on small coalition partners, which would make any government formed by the left less cohesive and more likely to collapse.Similarly, Bawn (1993) discusses this motivation in the Social Democratic Party's (SDP) preference for PR rules during the electoral reforms of 1949 and 1953 in Germany.
However, adopting a permissive reform also makes the parliament more accessible to all small parties, including the ones ideologically closer to the incumbent.For this reason, the governing party will be reluctant to implement such an electoral reform if its own cluster is already fragmented.In other words, this strategy would be beneficial only when the incumbent can afford it.This argument leads to the following hypothesis: Hypothesis 2: The higher the number of parties faced by the opposition relative to the governing party, the more likely permissive reform attempts are.

Data and methodology
I test the hypotheses using a novel dataset of party competition and electoral reform attempts that covers 32 parliamentary democracies from 1945 to 2015.Only parliamentary democracies, where the government formation and survival depend on the legislative elections, are included in this article, as the explanation here considers the calculations of parties about government formation in future elections.The resulting sample consists of 32 countries and 431 legislative terms in the dataset. 7The unit of analysis is the country-legislative term since most independent and control variables are based on the election results. 8The dependent variable is an electoral reform attempt, defined as an attempt to change one or more of the four main components of the electoral system affecting its proportionality: district magnitude, electoral formula, tier structure and legal thresholds.The attempts consist of the proposals drafted both by the government (and submitted to either parliament or referendum vote) and by a committee appointed by government (see also Shugart 2008).
The dependent variable takes three values: a permissive attempt, a restrictive attempt, or no attempt.Permissive reform attempts aim to make electoral systems more proportional and inclusive for new or small parties, whereas restrictive ones do the reverse. 9I identified a total of 121 electoral reform attempts; 42 of them are in the permissive and 79 are in the restrictive direction. 10Of these reform attempts, 33 permissive and 36 restrictive ones were successful.This quick glance at the data shows that we would observe a similar number of permissive and restrictive reforms if we only considered successful reforms in the study, which would introduce a bias in the empirical analysis.
In some countries, governments initiate more than one reform attempt within a legislative term, but since the unit of analysis is the country-legislative term, only one reform attempt is included in the regression analyses.Another option would be to change the unit of analysis to country-year instead of country-legislative term.However, as the variation in the main independent and control variables is based on elections, this approach would not be ideal.Moreover, the country-year approach is still limited; it does not consider multiple reform attempts in a given year.Therefore, the analyses include the last reform attempt in the same legislative period in such cases.For example, the Turkish government initiated two permissive reforms in 1995.The first one was more comprehensive; it added a national tier of 100 PR seats in addition to increasing the average district magnitude.However, this reform was found to be unconstitutional by the Constitutional Court for introducing a national tier.The government, then, eliminated the national tier component from the electoral law, successfully enacted the reform that increases the average district magnitude (Issever-Ekinci, 2016).Another example would be the Israeli electoral reform of 1973.Throughout the 1950s and 60s, Israeli governments wanted to make their electoral system more restrictive.In 1971, one such attempt was to decrease the district magnitude by introducing multiple electoral districts to their nation-wide PR system.When this attempt did not find support after the first reading in the parliament, the government implemented a restrictive reform on a smaller scale and changed the electoral formula from Hare to d'Hondt (Rahat 2008: 66-69).As these examples show, the difference between the reforms within the same legislative term lies in the components of electoral systems, affecting the magnitude rather than the direction of electoral reform.Therefore, the inclusion of only one attempt in the analysis did not affect the estimations about permissive and restrictive reform attempts.All in all, this coding yields a total of 106 reform attempts; 42 of them are permissive and 64 of them are restrictive.
In order to test the hypotheses, I develop a novel party system measure, the effective number of parties in ideological clusters, by considering the spatial distribution of political parties on the left-right dimension.I use Parliaments and Governments dataset on cabinets, elections and parties (Döring and Manow 2018) to calculate this variable as follows.First, the main government party (the party with the highest vote share) and the main opposition party (the runner up) are identified in an election.Then, other parties are categorised according to their ideological distance from the main government party and the main opposition party.That is, if Party A is closer to the party in government, then Party A is included in the government party's ideological bloc.The rationale here is that voters of Party A are potentially more willing to vote for the party in government than the main opposition party.After assigning parties to the ideological blocs of main party government and opposition, the vote share of each party is weighted by the total vote of parties in a given bloc, except for the main parties.Finally, the effective number of parties for the governing party (and for the main opposition party) cluster is calculated where the vote share of each party is normalised by the total vote received by all parties in the same ideological bloc.
To give the precise formula for calculating the competition in each bloc, let N G denote the number of parties belonging to an ideological bloc G in an election.To reweight the vote share of each party in bloc G, I first compute the total vote share of bloc G by summing up vote share of each party in this bloc, p p The same formula is used for calculating the effective number of parties in the opposition's ideological cluster. 11 I use several control variables in the empirical analyses.First, I control for the disproportionality of the existing electoral system by using the least square index of elections by Gallagher (1991).I also include the effective number of parties in the elections by Laakso and Taagapera (1989).Second, I control for the electoral volatility, using the Pedersen Index to calculate both new party volatility and within system volatility (Pedersen 1979).In some models, I control for the satisfaction with the electoral system.Pilet and Bol (2011) show that parties dissatisfied with the electoral system, i.e. parties that have not been in government for the last 25 years, are more likely to support an electoral reform.One of the implications of this finding is that if parties in government change frequently, governing parties would be more willing to implement an electoral reform to benefit themselves. 12 Another control variable is economic performance, measured as the annual change in the real GDP per capita.The literature suggests that the economic performance of governing parties can affect their electoral fortunes, and by extension, their decision over electoral reforms.For instance, poor economic performance during a legislative term might lead the incumbent to change the electoral system.I also control for new democracies, as studies suggest that they experiment with electoral systems more frequently than mature democracies (Bielasiak and Hulsey 2013, Remmer 2008, Renwick 2011).Lastly, I control for constitutional protection which measures the extent to which constitution protects the electoral systems; and government type which is a dichotomous variable for the single-party majority government. 13The descriptive statistics of the independent and control variables are shown in Table 1.
In order to estimate the likelihood of electoral reform types, I use the multinomial logistic regression since the dependent variable consists of three unordered categories.Multinomial logistic regression fits separate binary logits for each pair of outcome categories simultaneously, and therefore, estimates parameters more efficiently in comparison to modelling each outcome (permissive and restrictive reform attempts) with separate logistic regressions (Long and Freese 2014).In the regression models, the baseline outcome is no reform attempt, meaning that the coefficients in the permissive and restrictive attempt panels are the comparisons with the no attempt outcome.To account for the time dependence, I employ two strategies.First, I include a variable controlling for the number of previous events, i.e. the number of previous reform attempts, as suggested by Beck et al. (1998). 14Second, I report standard errors clustered at the country level to account for the hierarchical nature of the data.

Findings
Given that failed reforms are tilted towards restrictive attempts and that it was not captured in the datasets used by existing studies, I first discuss the findings regarding the earlier explanations reported in Table 2.Note that this is not a replication of their findings, since earlier studies used a different dependent variable: successful reform types.I revisit these explanations to assess the extent to which they help understanding of electoral reform attempts.In Model 1, the variable of interest is electoral volatility.The results show that electoral volatility is positively related to both permissive and restrictive attempts controlling for the economic performance, being a new democracy, disproportionality, constitutional protection, competitiveness of elections, and government type.This finding suggests that governments are more likely to initiate an electoral  notes: standard errors in parentheses are clustered at the country level.*** significant at the 1 percent level; ** significant at the 5 percent level; * significant at the 10 percent level.
Models 2 and 3 introduce the effective number of parties in an election, which measures the degree of party system fragmentation, to assess whether electoral reforms are initiated to reinforce or reverse the existing trends in the party system.Party system fragmentation has a negative relationship with permissive reform attempts, and a positive relationship with restrictive attempts.However, only the latter relationship is statistically significant.In other words, this finding lends partial support to the second explanation that electoral reforms reverse the existing trends in the party system while not supporting either explanation with respect to their predictions on permissive reforms.Model 3 controls for electoral volatility and the coefficient on the number of effective parties is reduced dramatically and it is not statistically significant anymore.Hence, the support for the second explanation is not only partial but also weak.
Models 4 and 5 introduce different types of electoral volatility to assess whether restrictive attempts are initiated to reverse the vote shifts in elections stemming from new parties.The results show that new party volatility increases the likelihood of restrictive reform attempts, lending support to the cartelisation mechanism suggested by Nunez et al. (2017).However, the results also show that vote shifts between existing parties, within system volatility, have also a similar effect on the restrictive reform attempts, posing a sharp contrast to the argument.In sum, electoral volatility is an important determinant of electoral reform attempts, regardless of its source and type of reform attempts.Overall, none of the potential explanations for permissive reform finds empirical support in this analysis.
Table 3 reports the findings of multinomial logistic regression analyses testing Hypotheses 1 and 2. In the baseline model (Model 1), I control for constitutional protection new democracy, economic performance, disproportionality, government type and competitiveness of elections.The results from this model provide support for Hypothesis 1: fragmentation in the governing party's ideological cluster has a negative and statistically significant (at the five percent level) effect on permissive reform attempts, and a positive and statistically significant effect (at the one percent level) on restrictive attempts, holding other variables constant.To get a sense of the magnitude of the main effect, consider the results from the baseline model with relative risk ratios. 15Accordingly, if the number of parties in the governing party's ideological cluster were to increase by one, the relative risk of observing a restrictive reform attempt, as opposed to not observing any attempts, would be expected to increase by a factor of 1.288, holding other variables constant.On the other hand, if the number of parties in the government's ideological cluster were to increase by one, the relative risk for initiating a permissive reform would decrease by a factor of 0.770. 16The results also show that new democracies are more likely to initiate both permissive and restrictive reforms and this positive effect is statistically significant (at the five percent level).The economic performance has a negative and statistically significant (at the one percent level) relationship with permissive reform attempts, suggesting that permissive reforms are less likely when there is an upward trend in GDP per capita.But it does not have a statistically significant relationship with the restrictive attempts.Conversely, the constitutional protection has the expected sign for both types of reform attempts, but its effect is only statistically significant (at the five percent level) for the restrictive ones, indicating that governments are less likely to initiate restrictive reforms as the constitutional protection of electoral system increases.Neither the disproportionality nor the competitiveness of elections has a significant effect on permissive and restrictive attempts.Lastly, the effect of being a single-party government is positive but not statistically significant for either permissive or restrictive reform attempts. 17 In order to assess the robustness of the main finding, I first add the satisfaction with the electoral system in Model 2. Measured as a dichotomous variable indicating if the major governing party took part in the government during the previous legislative term, the satisfaction with the electoral system has a negative relationship with both permissive and restrictive reform attempts.However, its effect is only statistically significant (at the one percent level) for restrictive attempts, indicating that incumbent parties are less likely to initiate a restrictive reform, when they are satisfied with the electoral system in use.Conversely, the negative effect of satisfaction is not statistically significant for either permissive or restrictive reform attempts when it is measured in the longer term (i.e.whether the incumbent party partook at least once in the last two legislative terms).The interesting conclusion that can be drawn from these results is that parties evaluate the benefits of electoral systems in the short term. 18 In Model 3, I include electoral volatility as a control variable.Consistent with the earlier findings, electoral volatility has a positive and statistically significant relationship with both permissive and restrictive reform attempts (at the five percent and one percent levels respectively).Importantly, the result concerning the main variable of interest does not change; fragmentation in the incumbent's ideological cluster has a negative and statistically significant effect on permissive reform attempts and a positive and statistically significant effect on the restrictive ones.Lastly, while adding electoral volatility does not eliminate the effect of fragmentation in the governing party's ideological cluster, the effects of being a new democracy and constitutional protection disappear in Model 3.
I replace total electoral volatility with new party volatility in Model 4. The relationship between the fragmentation in a government's ideological cluster and reform attempts does not change.When fragmentation in the government's ideological clusters is high, incumbent parties are less likely to initiate a permissive reform but they are more likely to initiate a restrictive reform, even after controlling for new party volatility.In fact, replacing total volatility by new party volatility increases the magnitude of the main effect of interest for restrictive reform attempts.As a result, the coefficient on the fragmentation in the government's ideological cluster becomes statistically significant at the five percent level.This suggests that a part of the effect of volatility on the likelihood of restrictive reform attempts is captured by the degree of fragmentation in the governing party's ideological cluster. 19 As these last two models demonstrate, fragmentation in the governing party's ideological cluster still has a substantial effect on restrictive reform attempts even after controlling for different electoral volatility measures.Holding other variables constant, given one unit increase in the fragmentation variable, the relative risk of observing a restrictive reform attempt, as opposed to not observing any attempts, would be expected to increase by 1.205 in Model 3 and by 1.235 in Model 4. On the other hand, as a response to a one unit increase in the governing party's cluster, the relative risk of observing a permissive reform would be expected to decrease by a factor of 0.755 in Model 3 and 0.771 in Model 4.
Figure 1 below shows the substantive effect of the fragmentation in the governing party's ideological cluster by presenting its marginal effect on permissive and restrictive attempts, holding the other variables at their mean values, and using the coefficient estimates from Model 4. When the effective number of parties in the government's ideological cluster is low, the probability of restrictive reform attempt is close to ten percent, holding other variables at their mean.As the governing party's competitors grow in number, the probability of a restrictive attempt starts to increase as well, and when the effective number of parties in a government's cluster reaches the maximum (around seven), the probability of initiating a restrictive reform approaches thirty percent.As for permissive reform attempts, when the government's ideological cluster is not fragmented, its probability is around fifteen percent.As the effective number of parties in the government's ideological cluster increases, the probability of permissive reform attempts starts to decline rapidly.
In Models 5 and 6, I test Hypothesis 2 which postulates that permissive reforms are more likely to be initiated if the opposition party faces more competition than the governing party.To test this hypothesis, I use the difference between the (effective) number of parties faced by the opposition party and those faced by the governing party.I find that governments are more likely to initiate permissive reforms if fragmentation is higher in the main opposition party's cluster than that of the incumbent.This effect is statistically significant at the ten percent level.Consistent with the set of results discussed earlier, new democracies are more likely to initiate permissive reform attempts.In addition, governments are more likely to initiate permissive reforms when there is a downward trend in their economic performance.According to these results, when there is a unit increase in the difference between fragmentation in ideological clusters, the relative risk of observing a permissive reform attempt as opposed to not observing any attempts would be expected to increase by a factor of 1.130, holding other variables in the model constant.When the government's ideological cluster is more fragmented than that of the opposition, the probability of a permissive reform is close to 5 percent.When they have an almost equal number of competitors, the probability of permissive reform increases to ten percent, and it continues to increase as the opposition's cluster becomes more fragmented. 20However, this relationship is not robust to the inclusion of electoral volatility in Model 6.So, there is weak support for the second hypothesis.
To explain how the party competition works during an actual reform process, I examine two typical cases of restrictive and permissive reform attempts selected according to the regression analyses.The first restrictive attempt case is the Lithuanian electoral reform of 2000 which changed the nominal tier of the Lithuanian parallel system from majority to plurality formula.The Homeland Union (Conservatives) government introduced this reform in the context of increasing the number of parties in its ideological cluster.The party system was in flux with party splits, important figures switching parties and new parties emerging.The Conservatives suffered the most from the new developments in the party system: MPs splitting from the Conservatives formed the Union of Moderate Conservatives.In addition, former PM Paksas resigned from the Conservatives and later became the new chairman of the Liberal Union Party.Moreover, the local elections that took place in 2000 signalled that the Conservatives and their allies would suffer from the vote shifts (Clark and Prekevičius 2003).Realising that new and small parties are more successful in the nominal tier, the Conservatives wanted to impede their potential success (Martinaitis 2012).The local elections also showed that the Conservatives were not able to win the mayorship in places where they had the highest vote share.Therefore, replacing the majority rule with the plurality formula would help the Conservatives (Clark and Prekevičius 2003).Introduced by the coalition government three months before the 2000 election, this reform met resistance from the opposition parties and was vetoed by the President.Yet the governing coalition garnered 72 votes in parliament to override the veto and alter the electoral system (Lithuanian parliament amends election law 2000). 21 Similar conditions also existed during the Latvian electoral reform attempts in 1995 and 1997.In both episodes, the fragmentation in the government's ideological cluster was high (4.6 and 5.6 respectively).The first attempt was successful in increasing the national election threshold from 4 to 5 percent, whereas the second one failed to introduce an electoral threshold for pre-election coalitions before the 1997 election.Both reform attempts, though, were proposed within the context of changing party system dynamics.When the Latvian Way-led (LW) coalition government proposed to increase the electoral threshold, several new parties such as For Latvia, the Christian Socially Conservative Popular Movement, and Saimnieks (Master) emerged.The latter two soon merged to form the Democratic Party-Saimnieks (Latvian Update 1995; Pettai and Kreuzer 1998 p: 156).Even though the opposition parties also drafted different electoral law proposals, the LW was able to enact its own proposal successfully in May 1995 (Latvia Update 1995b.)The new electoral threshold spurred a process of forming pre-election coalitions before the election and small parties entered the parliament through the electoral coalitions, rendering the new law useless.In fact, no parties or blocs were able to claim a parliamentary majority and the government formation proved so difficult that a surplus government led by a non-affiliated prime minister was formed (Sprudzs 2001).The leading parties in government, this time, wanted to reduce the proportionality of the system by introducing a 7 percent threshold for pre-election coalitions and a 3-year registration requirement to participate in elections (Latvia Update 1997a: 20-21).Given the fractured nature of the parliament and dependence of parties' seat share on the pre-election coalitions, leading parties did not find support for this reform in the parliament.
In contrast to these restrictive attempts, the permissive reform attempts in Lithuania and Turkey display different dynamics.To start with, the Lithuanian electoral reform of 2004 was introduced by the coalition government of the Social Democratic Coalition and Social Liberals to reverse the earlier reform, bringing back the majoritarian formula for the nominal tier.Before this reform attempt, the effective number of parties in the government ideological cluster was low (equalling 1) whereas the opposition's ideological cluster was much more fragmented.However, the major change before this reform was the emergence of the Labour Party which pursed a populist agenda and had anti-establishment election pledges (Duvold and Jurkynas 2012: 129).The Labour Party quickly became very popular and was ahead of parties in all polls before the European Parliament elections (Jurkynas 2005;Martinaitis 2012).With this reform, the coalition government targeted mostly the Labour Party whose support was very strong in the rural areas and periphery (Jurkynas 2005).In fact, the 2004 election results showed that the Labour Party was the first runner in 48 electoral districts in the nominal tier but was able to get only 16 of them (Krupavicius 2005).
Similarly, the coalition government in Turkey introduced a permissive reform of 1995 when the effective number of parties in the government's ideological cluster was low (1.05) and the opposition party faced more competition.The push for electoral reform emerged after the results of the local elections in the previous year.Even though True Path Party, the leading party in government, had still the most votes in the elections, the other two right-wing parties also had similar vote shares.Welfare Party, a relatively small party in earlier elections, emerged as the third party with the most efficient vote distribution (Issever-Ekinci 2016).Moreover, the support for the True Path Party-led coalition was further declining due to the economic crisis.In this context, the government introduced a proposal which enlarged the district size to increase the average district magnitude and introduce a national tier of 100 seats to be elected with d'Hondt formula.When the use of a national tier was found unconstitutional by the Constitutional Court, the government kept the basic features of the previous bill while eliminating the national tier. 22 This electoral bill was successfully ratified and used in the 1995 election.
The permissive reform cases uncover some of the dynamics observed in the regression models: they are introduced by the governments whose economic policies performed poorly, and thereby which expect to lose in the elections.The cases also demonstrate that the governing party is motivated by the emergence of a rising party, to initiate a permissive reform.Importantly though, the governing party did not have high numbers of parties in its ideological cluster, therefore, it was safe to open up the system.It is worth noting that the decision process for permissive reforms-the idea that the emergence of challenger parties incentivizes the government for such reforms-is also consistent with the argument put forward by Colomer (2004) and Remmer (2008).However, their suggested measure, the effective number of parties, is not sufficient to capture the dynamics explored in the two cases analysed above since it does not consider whether newly formed parties are ideologically close to the governing party or not.

Discussion and conclusions
Recent literature on electoral reform has shown the various ways electoral systems are altered, diverging from the earlier approach which emphasised the rarity of electoral reforms.Adopting a slightly broader definition of electoral reforms, these studies explored how the electoral context, such as party system fragmentation and electoral volatility, shapes the incentives of political parties in adopting different types of electoral reform.However, the existence of many electoral reform initiatives which failed in the process requires us to revisit the explanations about when and why politicians decide to change electoral systems.
In this article, I proposed a more nuanced explanation about the electoral context in which permissive and restrictive electoral reforms are initiated.Instead of focussing on the overall instability in the party system, I emphasise that party system fragmentation might have an asymmetric effect on the established parties; since smaller or new parties can draw votes away from one of the leading parties, hurting its electoral prospects and shaping its incentives to reform.Employing a novel dataset of party competition and electoral reform attempts in 32 democracies from 1945 until 2015, I have found support for the hypothesis that governing parties are more likely to initiate restrictive reforms when there are high numbers of parties drawing votes from them, whereas they are less likely to initiate permissive reform under the same condition.Relatedly, governing parties tend to initiate permissive reforms, which lower the threshold of representation, to weaken the opposition.But they do so only when it is safe for them to open up the system, in other words, when they face lower fragmentation in their ideological cluster than the leading opposition party.
This research makes two contributions to the broader literature on electoral system change.First, this article further qualifies the main conclusion about the rarity of the electoral system changes: the observed stability is an artefact of inattention to many reform attempts that fulfil the narrow definition of electoral reform but failed in the process; as well as many that succeeded but did not fit the criteria of electoral reform.While recent literature pointed to the latter category and broadened the definition slightly (Bielisak and Hulsey 2013;Nunez andJacob 2016, Nunez et al. 2017), I emphasised the former group and the need to include failed reform attempts by governments to understand when and why the party in power wants an alternative system.Indeed, I show that the inattention to unsuccessful reforms censors many restrictive reform attempts that fail and gives a roughly similar number of permissive and restrictive reforms.In other words, restrictive reform attempts are more likely to fail than permissive ones.Relatedly, the analyses also contribute to the understanding of the constraining capacity of institutions: parties in power refrain from initiating a restrictive reform when the constitution protects the electoral systems.These findings altogether point to the need to recognise different capacities that governments have in initiating and implementing different types (permissive and restrictive) of electoral reform.
Another theoretical implication of these findings is that parties in government are concerned about how the changes in the electoral arena can affect their future electoral and legislative power as well as the power of their major competitor in deciding to revise the electoral system.The literature on party system stability does not necessarily identify the political actor who decides to change the electoral system; but instead hints that large or established parties coordinate to change the electoral system (Nunez et al. 2017).However, case studies of electoral reforms emphasise that large parties do not necessarily have a similar interest.For example, various electoral system changes in Greece were motivated to keep the second largest party seats low by either giving premiums to the largest party or strengthening smaller parties at the expense of the second party (Dinas 2020).Bawn (1993) also demonstrates that the largest two parties in Germany supported different electoral systems during the electoral reforms of 1949 and in 1953.I applied this insight to dynamically model electoral reform attempts to predict which types of electoral reform are initiated by the parties in government.This study is a first attempt to explain electoral reform attempts by focussing on the governing party's incentives through providing a measure to identify whether it is disadvantaged by the party system fragmentation or not.In doing that, I use the proximity of small and new parties to the two largest parties in the ideological spectrum and generate a new variable: fragmentation in the ideological cluster of the governing party and its main competitor.
The ideological position of new parties has been neglected in the literature on electoral reforms in contemporary democracies, despite its central role in the explanations of PR reforms in the early 20 th century.This study brings back the ideology of the emerging parties to assess from which one of the largest parties they draw votes and develop electoral scenarios based on that.Therefore, it differs from the literature on PR reforms in two important ways.First, the new party for which the institutional solutions were sought was the left parties in the early 20 th century since the suffrage extension led to the massive increase in mostly left-wing voters.The only variation among the new parties in terms of ideology was based on their radicalism (Ahmed 2010;Ahmed 2013).In contrast, the new and small parties in contemporary democracies, emerge on both sides of the ideological spectrum.Second, I consider scenarios under which a permissive or a restrictive reform is attempted.The rule changes in the early 20 th century were always in the permissive direction.As Ahmed (2010:1069) points out, even when the plurality rule in single member districts was adopted in the United Kingdom in 1884, it was in the permissive direction because the initial rule was the plurality in multi-member districts.In contrast, the contemporary reforms can be in either the permissive or restrictive direction.
Nevertheless, the proliferation of anti-establishment parties across Europe raises the question of whether these parties pose a comparable risk to the way democracy works and whether and how parties in government seek institutional solutions to contain them.One way to study this question is to focus on the radicalism of these small and new parties, and to analyse whether it increases the likelihood of electoral reform attempts, regardless of their position on the left-right dimension.Further research is needed to explore how the radicalism of new parties is associated with permissive or restrictive reforms.
Finally, adopting the current model to majoritarian systems would be a fruitful avenue for further research.While the asymmetrical effect of party system fragmentation is also relevant for the majoritarian systems, the institutional response of the disadvantaged party might differ from that in the PR systems.Existing work already emphasises the resilience of majoritarian systems in the post-war era, given that the will and power to change the plurality systems are reversely related (Blau 2008;Mitchel 2005).However, the emergence of new parties or increase of minor parties in size can create asymmetrical benefits for major parties, hurting one of them more than the other.Further research is needed to understand when these asymmetrical effects create an incentive to change the electoral system and which types of reform are initiated in the majoritarian systems.2. Incumbent party is the political party currently holding office in government.I use the terms incumbent party, governing party, and party in power interchangeably to convey the same meaning throughout the article.3. See also Blais (2008) which examines electoral reforms from the First Past the Post System; Grofman (1999) and Bowler and Grofman (2000) which focus on the Single Non-Transferable Vote and the Single Transferable Vote respectively to explain their origins and effects.4. Some options such as increasing the district magnitude in plurality systems are obsolete and not used in any consolidated democracies.Others, such as decreasing the assembly size, are difficult because the parties need to convince the backbenchers to give up the seat in their district.Finally, while boundary delimitation can produce partisan, and in particular, incumbent advantage, it does not necessarily affect the proportionality of the system. 5.A similar assumption is also used by Bawn (1993).6.The argument does not assume that there are two strong parties, either of which forms a government in each election.It allows for the possibility that the main parties in government and an election can differ depending on the election results, and that the main parties in government and an election can have an imbalance in terms of electoral power.7.For the list of countries see online appendix Table 1 and for the number of legislative terms each type of reform is attempted in the countries see Figure 1 in the online appendix.8. See also Nunez et al. (2017), Bielasiak andHulsey (2013), Remmer (2008).9. See online appendix Table 2 for the coding procedures and sources for the dependent variable.10.Two electoral reform attempts were not included in the dataset as the direction of change was not clear: i) German reform (2005) which replaced Hare quota with the Sainte Laguë; ii) Romanian reform (2008) which introduced single member candidacies yet distributed party seats as in the previous system (Marian and King 2010;Renwick 2011).11.The main parties are not included in the calculation of competition in each bloc with the aim to derive a fragmentation score sensitive to the small parties.12.I thank the reviewers for their suggestion to include the satisfaction with the electoral system in the analyses.13.See online appendix Table 3 for the definition and data sources for each control variable discussed here.14.The number of previous reform attempts is included in all regression models in this article but not reported in the tables for space considerations.In each model this variable is close to zero and statistically not significant.For similar approaches used in multinomial regressions to deal with the time dependency problem, see Bielasiak and Hulsey (2013); Nunez and Jacobs (2016); Nunez et al. (2017);Remmer (2008).15.Note that relative risk ratios are exponentials of multinomial logit coefficients, and they indicate the risk of a permissive (or restrictive) reform attempt in comparison to no reform attempt by a unit increase in the independent variable.16.Note that when the relative risk ratio is below 1, the outcome is more likely to be in the baseline group (Long and Freese 2014).17.I also estimated each specification in Table 3 using the presence of judicial review as an additional control variable.The qualitative nature of the results did not change.Because the coefficient on judicial review was not significant in any of the estimated models, I do not report those results.18. See the results in online appendix Table 4.I also consider the possibility that the relationship between fragmentation in a government's ideological cluster and the electoral reforms might differ depending on the satisfaction of the incumbent with the electoral system.In Models 1 and 3, I test this interactive effect.The results show that the effect of fragmentation in a government's ideological cluster on permissive and restrictive reform does not depend on its satisfaction with the existing system.19.The effect of fragmentation in the governing party's ideological cluster is also robust to the inclusion of fragmentation in the opposition's ideological cluster variable (models not shown here).20. Figure not shown for space considerations.21.The electoral reform, though, benefited the leftist coalition the most in the 2000 election only because they were able to form a pre-election coalition.22.The permissive reform makes the electoral system more proportional but only for the parties that pass the infamous 10 percent national threshold.
Next, I compute the reweighted vote share of each party belonging to the ideological bloc G, p n  , as p p p n n G  = / .In the last step, I compute the effective number of parties in ideological bloc G using the following formula: Effective Number of Parties in bloc G

Figure 1 .
Figure 1. the effect of fragmentation in governments' ideological cluster on reform attempts.

Table 2 .
analysis of existing explanations in the literature.

Table 3 .
Fragmentation in government's ideological cluster & electoral reform attempt.
notes: standard errors in parentheses are clustered at the country level.***significant at the 1 percent level; **significant at the 5 percent level; *significant at the 10 percent level.