Abstract
This article analyses the causal effect of the 1993 electoral reform in New Zealand on party system fragmentation using the ‘synthetic’ control method. Previous studies using cross-national evidence suggest that electoral reforms change the number of parties. However, they do not take into account possible endogeneity problems and usually focus on their short-term effects. Since the electoral system in use in this country before the change was first past the post (FPTP), I can create a ‘synthetic’ control democracy that had the same institutional framework but did not modify the rules of the game. The results indicate that the electoral reform produced the expected effects on party system size at the electoral level, but that these effects tended to disappear in the long run. In contrast, electoral system effects at the legislative level were larger and stickier over time.
Major electoral reforms are believed to have strong effects on party system fragmentation. The field in general takes as an axiom that electoral system changes modify the number of parties by affecting the translation of parties’ votes into seats and the strategic calculations of voters and elites. However, empirical works on this matter that employ appropriate research designs are still scarce. A key reason for the lack of causal evidence in this regard is that electoral reforms are not assigned randomly. Important contextual characteristics influence the adoption of electoral reform, which in turn affects the likelihood of party system change. This article investigates the hypothetical impact of electoral reform in first past the post (FPTP) systems by using the switch to proportional representation (PR) in New Zealand during the 1990s as a case study and employing the synthetic control method to address possible endogeneity problems.
New Zealand constitutes a perfect case study, because FPTP is the most frequently employed electoral system in the world. More specifically, the New Zealand episode is especially interesting for three reasons. First of all, major electoral reforms like this one are fairly uncommon in established democracies (Lijphart, 1994). Hence, if we want to know the potential effects of the actual abandonment of FPTP at the national level for the party system, we need to focus on the New Zealand reform. Secondly, the electoral system in this country, in contrast to other cases of electoral reform, has not changed again after 1993. Consequently, institutional instability in New Zealand after this year is unlikely to have had a substantial mitigating effect on the impact of the original electoral reform. More importantly, the pre-reform New Zealand system was described by comparative scholars as ‘a virtually perfect example of the Westminster model of democracy’ (Lijphart, 1984: 16). In the period of hegemonic two-party politics that began in 1938, the effective number of electoral parties rarely reached three. 1
Yet, this pattern started to change radically in the mid-1990s and early 2000s, with an effective numbers of electoral parties of at least 3.5. During this period, the number of parties that ran in elections grew, and arguments by the leaders of the major parties about wasted votes did not deliver a reduction in party system fragmentation. Not surprisingly, the change registered in New Zealand politics has been attributed, at least partially, to the effect of electoral reform. However, little research has been carried out to assess the causal impact of the electoral rules change on the party system, because examining the effects of electoral reforms usually involves two big difficulties.
On the one hand, a pure time-series analysis of the institutional change and the evolution of the New Zealand party system would be contaminated by the inflation in the number of parties registered in other established democracies after the erosion of the traditional social cleavages described by Lipset and Rokkan in 1967 (Franklin et al., 1992). On the other hand, when the new electoral rules were adopted, New Zealand differed from other FPTP countries in characteristics that are thought to be related to party system fragmentation. Therefore, a simple comparison of the evolution of the New Zealand party system and the party system of the rest of FPTP democracies would reflect not only the effect of the electoral reform but also the effect of pre-reform differences in determinants of the number of parties.
In order to tackle these problems, the synthetic control method is employed. According to Abadie et al. (2015), this approach is useful to complement and facilitate comparative cases studies in political science and allows us to know how the number of parties in New Zealand would have evolved in the absence of electoral reform. More specifically, the analysis rests on the following identification strategy: I use a combination of other FPTP countries to construct a synthetic control majoritarian democracy that resembles relevant political characteristics of New Zealand before the adoption of a PR system. The subsequent party system fragmentation of this ‘counterfactual’ New Zealand without reform is compared to the actual experience of New Zealand. Doing so, it is possible to isolate the net effect of electoral system change, circumventing the problems of selection associated with the economic reforms of those years in New Zealand and, above all, the process of enactment of the new electoral institutions. As in extant empirical literature, observational data will be also used here, but I will adopt a sophisticated research design to avoid endogeneity issues and validate or refute as a consequence previous results. The research design adopted also allows studying the short- and long-term effects of electoral reforms at the same time.
The article now proceeds to review previous works on the effects of electoral rules on party systems. In the second section, I summarize the history of the New Zealand electoral reform. I then elaborate on the research design adopted, presenting its important advantages, and overview the main results. Finally, I draw some conclusions.
Electoral rules and party systems
According to Duverger (1964 [1954], the negative consequences of restrictive electoral rules (mainly, single-member districts [SMDs] with plurality rule) on party system fragmentation are the result of two mechanisms. First, minor parties are typically awarded a much smaller share of seats than the share of votes they receive. Second, the existence of this mechanical effect creates incentives for electoral coordination by fostering, first, strategic withdrawal on the part of political entrepreneurs and, second, tactical voting on the part of voters (Cox, 1997). Duverger uses the term psychological to refer to these two behavioural consequences of non-permissive electoral laws for party system size. When the mechanical and psychological forces operate, district-level party systems will be characterized by two-party dominance (Crisp et al., 2012a; Hicken and Stoll, 2011). 2 In contrast, more permissive rules lead to larger party systems (Benoit, 2000; Blais et al., 2011).
Some previous empirical research on electoral systems has consistently found evidence in favour of the existence of a mechanical effect (Lijphart, 1994). In contrast, other authors show that electoral institutions only matter in the presence of sufficient social heterogeneity (Amorim Neto and Cox, 1997; Clark and Golder, 2006). Evidence of electoral systems’ psychological effects are similarly mixed; some studies find that electoral rules have an impact on the amount of strategic voting (Selb, 2012), while other studies claim that there is no evidence of the numbers of strategic voters being affected by electoral rules (Abramson et al., 2010).
The bulk of the literature on electoral systems has tested the mechanical effect by resorting to cross-national comparisons of countries with different institutional rules (Amorim Neto and Cox, 1997; Clark and Golder, 2006; Lijphart, 1994). A potential caveat to this literature is that part of the observed association between proportionality of the rules of the game and number of parties is thought to be provoked by reverse causation, since electoral permissiveness is not only a cause but also an effect of high levels of party system fragmentation (Colomer, 2005). In making this case, Colomer challenges the so-called Duverger’s Propositions (Grumm, 1958). The impact of party system configurations on the adoption of proportional representation at the beginning of the 20th century has been widely studied in the European context (e.g. Boix, 1999; Cusack et al., 2007). 3
Alternatively, electoral reforms became more frequent in some established democracies in the early 1990s, when popular discontent in Italy, Japan and New Zealand led to the adoption of a mixed-member electoral system in all three countries. Although all three were reforms by elite-mass interaction, because popular pressure on the majority of politicians forced electoral system change (Renwick, 2010), the electoral rules of origin varied widely and ranged from relatively disproportional – such as was the case in New Zealand – to considerably proportional – such as Japan and, above all, Italy. Moreover, the results of these institutional changes were not uniform across countries either and turned, for example, Japanese parties into more coherent and policy-oriented actors (Reed, 2005) or increased the complexity of government formation in New Zealand (Vowles, 2005). Interestingly, engaging with the literature on the Italian and Japanese reforms strengthens the case for using the synthetic control method to address the question of the direction of causality in New Zealand, because in all three countries the party system that had been remarkably stable for several decades began to unravel in the years before the reform (Bull and Newell, 1993; Lin, 2006).
Cross-national studies of the mechanical effects of electoral systems present a second problem, because they do not usually address the presence of unobserved heterogeneity that can drive the effect of electoral systems on the number of parties. For this reason, testing the within-country consequences of electoral rules has more recently become the focus of electoral system scholars (Giannetti and Grofman, 2011). Although the existence of mechanical effects has been demonstrated in within-country comparisons of different electoral arenas (Crisp et al., 2012b) or episodes of electoral system change (Riera, 2015), recent literature has been particularly prone to use evidence from PR and SMD tiers of mixed-member electoral systems (Moser and Scheiner, 2004).
Within this framework, causal inference techniques can also be applied to correct for endogeneity problems. Unlike existing studies that simply compare outcomes before and after reform to make inferences on electoral laws’ effects (Giannetti and Grofman, 2011), novel identification strategies allow researchers to approximate counterfactuals to serve as comparison units for countries that changed their electoral system. Despite its advantages, the use of causal inference techniques in the field is still in its infancy (for some exceptions, see Fiva and Folke, 2016; Fujiwara, 2011; Ward, 2015). This article complements previous findings, as it examines the effect of an electoral reform at the national level based on a comparison of actual and counterfactual seat allocation outcomes.
A second contribution of this piece is to analyse both the short- and long-term effects of changes in electoral institutions on party system size. As previously noted, many examinations of the relationship between electoral rules and the number of parties analyse it statically, and when they study the impact of changes in electoral institutions they do so in the short term (for a notable exception, see Best, 2012). By exploring the dynamic relationship between electoral rules and the number of parties in the long run, some evidence is found of how short-term increases in party system size at the electoral level provoked by a permissive reform may vanish after a given period of time.
New Zealand’s electoral reform and theoretical expectations
After more than eight decades of having consistently used the FPTP system for parliamentary elections, New Zealand decided to adopt mixed-member proportional (MMP) rules for the 1996 elections. Two popular referendums were necessary in order to override the preferences of major political parties and change the electoral system: an initial (non-binding) selection in 1992 from several options, followed by a later (binding) choice in 1993 between the status quo and the proposed change. This second consultation was co-scheduled with the 1993 general election. The former system had permanently favoured the two largest parties, since 1936, those being National and Labour, not only by underrepresenting the political parties outside the two dominant forces but also fostering tactical voting and abstention among the supporters of the latter. The new electoral rules were in turn supposed to help minor parties by raising their vote shares and increasing the number of MPs they got elected. Since their first use in 1996, there have been eight elections under the MMP rules. New Zealand voters can issue two votes: the constituency vote and the list vote, although it is the latter one that really determines the number of seats each party obtains most of the time.
The mixed-member system adopted in New Zealand is characterized by four basic aspects: first, the use of the plurality rule and the modified Sainte-Laguë formula in the nominal and list tiers, respectively; second, the presence of two alternative legal thresholds (either 5% of the national vote at the list tier or victory in one single-member race at the nominal tier) to be allowed to participate in the allocation of list seats; third, the existence of a single national PR district whose magnitude has ranged between 55 in 1996 and 49 in 2017; 4 and fourth, the possibility of increasing the theoretical size of the assembly – that is, 120 – with the so-called overhang seats when a party wins in more single-member districts than seats it is entitled to obtain according to its vote share at the list tier. 5
Previous analyses of this particular episode of electoral system change suggest that the adoption of the MMP system affected the political environment by fostering new parties’ entries (e.g. ACT or United New Zealand), old parties’ splits (e.g. the Green Party from the Alliance) and small parties’ vote shares (Vowles, 2005). It is obvious that the electoral reform operated by reducing voters’ incentives to abandon small parties at the list tier and enhancing the likelihood of either coalition or minority governments. Yet, it could be also argued that the presence of the 5% threshold contributed to the limited increase in party system fragmentation, as small parties found it difficult to surpass that threshold and voters were probably aware of it. In that context, some small parties needed to win at least one SMD to gain representation leading them to concentrate organizational efforts in a single constituency and weakening the incentives of their supporters in other districts to vote for them in the list tier.
Empirical strategy
Analytical methods and main results
The starting point of this research is that a simple comparison of party system fragmentation in New Zealand and other countries during the post-reform elections may not only reflect the impact of the institutional change but also other pre-reform differences which affected the subsequent number of parties. This problem is tackled by comparing the evolution of the New Zealand party system during the post-reform era with that of a weighted combination of other FPTP countries chosen to resemble the characteristics of New Zealand before the reform. I conceptualize such a weighted average of other FPTP countries as a ‘synthetic’ New Zealand without reform (Abadie and Gardeazabal, 2003; Abadie et al., 2015).
The synthetic control method presents important advantages relative to both small-sample qualitative and large-sample regression-type studies (Abadie et al., 2015). With regard to the former, the synthetic control method provides a systematic way to choose comparison cases and opens the door to quantitative inference. With regard to the latter, it makes explicit the contribution of each comparison unit to the counterfactual of interest, allowing, as a result, a more detailed analysis of the similarities and differences between the case of interest and the comparison units. Moreover, by making equal the histories of the control and the treatment units up to the intervention point, the potential problem of reverse causality largely disappears, because the synthetic control method assumes that pre-reform changes in the party system are the same in the actual and the counterfactual New Zealand.
To construct the sample, I build on previous works (e.g. Abadie et al., 2015) and start with the universe of contemporary democracies that employ a FPTP system so differences in the crucial feature of interest here (i.e. the employed electoral system) are minimized. Countries are included in the donor pool on the basis of two criteria: they must have used FPTP as an independent and democratic country since the mid-1970s 6 and they must have something that resembles a national party system. The reason for the first requirement is that the synthetic control method requires a balanced panel in which each unit is observed at every included time period. 7 Secondly, those countries where the extent of local and regional politics makes them qualitatively different are excluded from the donor pool, taking Papua New Guinea and the Solomon Islands out as a result. Thus, the FPTP cases employed to construct the synthetic New Zealand are the lower or sole house of the legislature in Antigua and Barbuda, Bahamas, Barbados, Botswana, Canada, India, Jamaica, Saint Vincent and the Grenadines, Trinidad and Tobago, UK and USA. 8
I consider five-year averages of covariates and electoral outcomes in each country as in other studies that use this technique (e.g. Abadie and Gardeazabal, 2003). Using five-year averages is necessary because, as pointed out above, the synthetic control method requires a balanced panel, and a panel of elections is far from balanced due to the different length of legislative terms across countries. The full list of elections employed is provided in Table A1 of the Appendix.
Following Laakso and Taagepera (1979), fragmentation is defined as the feature of party systems that takes into account the number of competing parties and their size, and is measured by using the effective number of electoral (ENEP) and parliamentary (ENPP) parties. 9 The former is typically used to characterize the psychological effects of electoral systems on voters, whereas the latter usually characterizes the mechanical effect of electoral systems (Blais et al., 2011). 10
To create the synthetic control that approximates the counterfactual New Zealand without reform, it is necessary to find covariates that explain party system size. These predictors should not be affected by the intervention, because otherwise it would not be possible to isolate the actual causal effect of the reform. Building on previous works on the determinants of party system size and opting for a parsimonious strategy as recommended by Abadie et al. (2015), the employed covariates comprise socio-demographic factors such as ethnic fragmentation in order to capture the effect of social cleavages on party system fragmentation and its potential interaction with electoral rules (Clark and Golder, 2006). 11 Federalism (Crisp et al., 2012a) 12 and democratic age are additional covariates (Tavits and Annus, 2006), 13 as well as assembly size and number of inhabitants (Taagepera, 2007). Finally, economic conditions are included as covariates as well. 14 I have opted for this operationalization in order to maximize the number of available cases for the analyses.
Table A2 in the Appendix reports the values of the variables typically associated with party system fragmentation that are used for the control FPTP countries (columns 1 and 5) and New Zealand (columns 2 and 6) during the pre-reform period. Relative to the whole sample, New Zealand has higher democratic age and economic growth but lower ethnic fragmentation. As can be seen in the table, the different versions of the synthetic New Zealand resemble better the features of the actual New Zealand than the average of all the countries. The third, fourth, seventh and eighth columns of Table A2 give an indication of how well the weighted combination of control countries reproduces the values of party system fragmentation predictors for New Zealand before the reform. Although the synthetic New Zealand looks more comparable to the actual one than the simple combination of the other FPTP democracies, most party system fragmentation determinants cannot be perfectly fitted, being particularly troublesome the fit with regard to ethnic fragmentation.
Table A3 in the Appendix shows the weights given to countries in the construction of the synthetic New Zealand in all analyses. This set of countries is heterogeneous and slightly varies across specifications. Some have similar socio-economic conditions to New Zealand, others have similar institutional structures. For example, if we examine the effective number of electoral parties and take the 1996 election as the treatment election (i.e. column 1 in Table A3), the optimal weights,
Despite the deficiencies with regard to how closely the synthetic case resembles the values of party system fragmentation predictors for New Zealand before the reform, Figure 1 shows that the synthetic New Zealand effectively replicates New Zealand’s pre-reform party system size. On average, the difference between the estimated party system for the synthetic New Zealand and the observed party system never exceeds 0.08 effective parties, and in the first analysis is only 0.01. In contrast, Figure A1 in the Appendix shows that before the 1990s New Zealand and the average FPTP country experienced different paths in the effective numbers of parties.
Party system fragmentation at the electoral and the legislative level for New Zealand and synthetic New Zealand.
The upper left panel in Figure 1 plots the party system fragmentation at the electoral level for the actual and the synthetic New Zealand for the whole period of observation, using the 1996 election as the treatment. Both cases behave similarly until 1996. From 1996 onwards, when the MMP system was used for the first time, the effective number of electoral parties in both scenarios diverges; the effective number of electoral parties in the treated case takes values up to around 1 above those of the synthetic control. The gap in party system fragmentation at the electoral level seems to decrease at the end of the period of observation. In fact, the effective number of electoral parties is actually lower for the treated case in the last three covered elections – that is, after 2008. However, Figure 1 suggests an average 0.5 electoral party increase in the effective number due to the reform for the period between 1996 and 2005.
The upper right panel in Figure 1 plots the party system fragmentation at the legislative level for the actual and the synthetic New Zealand for the whole period of observation, using the 1996 election as treatment again. Both cases behave even more similarly than in the previous analysis before the reform. From 1996 onwards, when the MMP system is already in use, the effective number of legislative parties in both scenarios diverges; the effective number of parliamentary parties in the treated case takes values up to around 1.5 above those of the synthetic control. The gap in party system fragmentation at the parliamentary level also seems to decrease at the end of the period of observation. This probably responds to the decline in the effective number of electoral parties we observed above. However, the effective number of legislative parties is never lower for the treated case. Even in the last covered election – that is, the 2014 election – this variable takes values up to around 0.5 above those of the synthetic control. Moreover, Figure 1 suggests an average increase of one legislative party in the effective number due to the reform for the entire post-electoral system change period. This is completely expected given the mechanical effects of electoral rules (Duverger, 1964 [1954]).
Bearing these findings in mind, the previous analyses are replicated by taking 1993 as the treatment election. Although MMP had not been formally implemented yet, it could be argued that the overwhelming vote for change in the non-binding 1992 referendum led politicians to expect that MMP would win in 1993. Anticipating the adoption of MMP, breakaway leaders were already positioning minor parties for post-MMP competition. Moreover, the second binding referendum on electoral reform was held in conjunction with the 1993 general election. This circumstance could have increased the incentives of supporters of minor parties to vote this time and, as a result, raise the effective number of electoral parties even though the new system was not yet in use. The evidence displayed in the panels at the bottom of Figure 1 seems to confirm these points. Although pre-treatment balance is slightly worse if we take 1993 as the treatment election, the growth in the number of electoral parties is already noticeable that year. Hence, we could argue that it is not the electoral system change per se but the calling of the referendum in conjunction with the 1993 general election, and the reform prospects that it brought about, that drove the effective number of electoral parties up. In contrast, results seem to confirm the outlined expectations about party system fragmentation at the parliamentary level: the growth in the effective number of legislative parties did not happen until 1996. The size of the estimated effect of the electoral reform in New Zealand on its levels of party system fragmentation can be seen even better in Figure 2, where the difference in this regard between the actual New Zealand and its synthetic version is displayed.
Effective number of parties gap between New Zealand and synthetic New Zealand.
Alternatively, two additional analyses are conducted to show that the results are consistent across different specifications. First of all, I take the effective number of electoral parties at the local (district) rather than the national level. This exercise presents the disadvantage that the number of countries where this indicator is available is considerably lower. For this reason, as Figure A2 of the Appendix shows, the synthetic control approximates party system fragmentation before the ‘intervention’ much worse. However, the main pattern stands, and we can see how the number of parties increases in New Zealand because of the reform and returns to its previous level after three elections. In contrast, as Figure A3 of the Appendix displays, the impact of the reform on the mechanical effect measured as the difference between the effective numbers of electoral and legislative parties is substantively more identical than the above and does not disappear over time.
Finally, the reader wonders what mechanisms could explain these fluctuations over time. Even though providing a well-developed discussion of the factors that drive the temporal sequence of the impact of the electoral reform would go beyond the main aims of this article, at least three empirical patterns seem to be responsible for the timing and duration of the effect. First, as Figure A7 of the Appendix shows, the percentage of respondents in the corresponding surveys of the New Zealand Election Study who, feeling they have a general attachment to a particular party, identify themselves with (declare themselves close to) neither the Labour nor the National Party peaks in 1996 and starts to decline afterwards. Second, Figure A8 of the Appendix shows that, only in the first decade after the adoption of the reform, the respondents who identified with a minor party were less likely to switch parties over elections. Before 1996 or after 2005 the effect is either positive or indistinguishable from zero. Likewise, according to Figure A9 of the Appendix, the effect of identification with a minor party on the likelihood of turnout is only consistently positive in the four elections that go from 1993 to 2002. In 1990 (because of the lack of viability of most minor parties’ candidates under the previous system) or after 2002 (probably because National and Labour keep winning elections despite the electoral reform), being identified with a minor party does not increase the likelihood of participating in elections.
Hence, despite how cursory this examination of the mechanisms of work is, we can conclude that party system fragmentation at the electoral level increases in New Zealand due to the reform process because those identified with a minor party are for some elections more likely to: 1) stick to their previous party choice; and 2) participate in elections. The latter effect is already present in the 1993 election. When these two patterns disappear (broadly speaking, from 2005 onwards), the effective number of electoral parties returns to its previous level.
Placebo study
Of course, a question remains about whether the gaps shown in Figure 1 truly respond to the effect of electoral reform or are merely an artefact of the inability of the analysis to reproduce party system fragmentation for New Zealand in the absence of reform. To address this question, a placebo study is performed, applying the same method that was previously used to compute the gap for New Zealand to a non-reform country (i.e. a FPTP system other than New Zealand). The idea is to see whether the gap observed for New Zealand may have been created by factors other than electoral reform. This placebo study is conducted by choosing Canada. 15
Canada was one of the countries with the largest weights in the synthetic controls for New Zealand according to the estimates displayed in Table A3. Hence, compared to the other FPTP countries, Canada matches New Zealand closer on the array of covariates employed to create the synthetic case. In addition to being the country most similar to New Zealand before the electoral reform in party system fragmentation determinants (as measured using these methods), Canada resembles New Zealand in some other characteristics, some of which are not directly measured in the data. In particular, a process of dealignment began in the 1960s in both countries (Dalton, 2012), which led to the weakening of the two major parties in each of them (i.e. the Liberal and the Progressive Conservative Party in Canada, and Labour and the National Party in New Zealand). The progressive decline of identification with mainstream parties took place in almost all established democracies at that time (Franklin et al., 1992); however, this trend came in conjunction with high levels of political discontent in New Zealand and Canada. While in both countries large fractions of the population demanded political reform, Canada never adopted a new electoral system.
Figure 3 shows the actual party system fragmentation path for Canada and the one implied by a synthetic Canada constructed as a weighted combination of the other FPTP countries (excluding New Zealand), as explained above. According to this figure, the weighted combination of FPTP countries reproduces party system fragmentation for Canada with slightly less precision than for New Zealand if we take 1996 as the treatment year (see the first row). The resemblance in pre-reform observations does not significantly improve if we take 1993 as the cut point (see the second row). In no scenario, the actual Canada outperforms the synthetic control after the treatment year. In fact, in the case of the effective number of electoral parties, party system fragmentation for the synthetic control is consistently higher after the simulated intervention than in actual Canada. In contrast, this gap does not exist when we look at the effective numbers of parliamentary parties.
A ‘placebo’ study, party system fragmentation at the electoral and the legislative level for Canada and synthetic Canada.
Since Canada is one of the main contributors to the synthetic control for New Zealand, an abnormally lower level of the effective number of electoral parties for the former during the years after the reform may artificially increase the party system fragmentation gap for New Zealand in Figure 1. Therefore, the placebo study suggests that, while pre-reform party system fragmentation can be reasonably well reproduced by these techniques, the effect of electoral reform on party system size at the electoral level for New Zealand from the 1990s onwards (relative to the synthetic control FPTP system) may have been slightly less pronounced than suggested by Figure 1. In order to tackle this problem, in the next section, a robustness check is run to test the sensitivity of the main results to changes in the country weights
Robustness test
Recall from the first column in Table A3 that the synthetic New Zealand in this specification is estimated as a weighted average of Canada, St Vincent and the Grenadines, UK and Trinidad and Tobago, with weights decreasing in this order. Here, I iteratively re-estimate the baseline models to construct a synthetic New Zealand in each case (one per each treatment year and dependent variable of interest) omitting in each iteration one of the countries that received a positive weight in Table A3. By excluding countries that received a positive weight, some goodness of fit is sacrificed, but this sensitivity check allows me to evaluate to what extent the obtained results are driven by any particular control country.
Figure 4 displays the results and reproduces Figure 1 (solid and dashed black lines) while also incorporating the leave-one-out estimates (grey lines). This figure shows that the results of the previous analysis are fairly robust to the exclusion of any particular country from the sample of comparison countries. The leave-one-out synthetic controls that show the smallest effects of the electoral reform are those relative to party system size at the electoral level that take 1993 as intervention year. Even these estimates are fairly large in substantive terms: The effective number of electoral parties still grows by about 0.5 units on average. This is an increase of approximately 25% of the 1990 baseline level. Hence, the other leave-one-out synthetic controls show either a very similar or a barely larger effect compared to the results of the previous analyses, giving us some additional confidence about the robustness of the main results.
Conclusion
In this article, the electoral system change adopted in New Zealand in the mid-1990s is used to examine the potential impact of electoral reform in other FPTP countries. Much has been said about the huge impact of electoral reforms on the features of party systems. However, to date little causal inference research has been produced on this matter. This article presents evidence of a positive impact of the adoption of MMP on the levels of party system fragmentation in New Zealand by, unlike other previous studies in the field, using sophisticated statistical methods.
In a nutshell, the conducted analyses suggest that, in the case of a permissive electoral system change in FPTP systems such as Canada or the UK, effects at the electoral level would be smaller and more short-lived than at the legislative level. More specifically, this study shows a 0.5 (1) electoral (legislative) parties average gap between New Zealand’s effective number of parties and the effective number of a comparable synthetic FPTP system without reform. Whereas the change in the effective number of parliamentary parties is expected given the mechanical effects of electoral rules, the impact on the effective number of electoral parties becomes remarkable if we take into account that the employed method minimizes the pre-reform differences and the size of the estimated effect is larger than the average fluctuation over the previous decade(s).
Leave-one-out distribution of the synthetic controls for New Zealand.
Yet, the estimated variation with regard to party system fragmentation at the electoral level shows two unanticipated patterns. First of all, the change in party system fragmentation at that level seems to emerge in the election immediately previous to the reform, casting as a result important doubts about the actual effect of the switch from FPTP to MMP. Following Colomer’s (2005) thesis, the rise of minor parties under a majoritarian system seems to precede the adoption of a permissive reform. Even more importantly, the modifications in the effective number of electoral parties tend to disappear after a number of elections conducted under the new electoral rules. Hence, we need to keep the same electoral laws for at least three elections before evaluating their full effects on voters’ behaviour (Taagepera, 2007). Otherwise, we will be overestimating their real effect.
Although the focus is on New Zealand, the methods applied here can be used to investigate the political effects of electoral reforms elsewhere. The application of the technique and the substantive ideas in this article to the study of changes in the rules of the game will also shed light on the robustness of the procedure and serve as cross validation. Even though this investigation has highlighted the importance of electoral reform for short- and long-term changes in the number of electoral and parliamentary parties, respectively, several tasks still need to be addressed in order to better understand the workings of this relationship. Within this framework, it would be important to assess whether electoral reform also affects other political outcomes (e.g. public expenditure). This article should spur the interest necessary to address some other questions and improve our understanding of how electoral reforms affect political outcomes, and to extend if possible these findings to other contexts and, above all, other reforms. Research of this sort could potentially have a decisive impact if policy makers learn that their actions have an effect on the levels of party system fragmentation.
Supplemental Material
IPS822891_French_Spanish – Supplemental material for Voting after the change: A natural experiment on the effect of electoral reform on party system fragmentation
Supplemental material, IPS822891_French_Spanish for Voting after the change: A natural experiment on the effect of electoral reform on party system fragmentation by Pedro Riera in International Political Science Review
Supplemental Material
IPS822891_supplemental_material – Supplemental material for Voting after the change: A natural experiment on the effect of electoral reform on party system fragmentation
Supplemental material, IPS822891_supplemental_material for Voting after the change: A natural experiment on the effect of electoral reform on party system fragmentation by Pedro Riera in International Political Science Review
Footnotes
Acknowledgements
An earlier version of this article was presented at the 2015 Conferences of the Midwest Political Science Association and the European Political Science Association. I would like to thank Elias Dinas, Dominik Duell and Bruno Pinheiro Wanderley Reis for helpful comments and suggestions.
Funding
This work was supported by the Spanish Ministry of Economy, Industry and Competitiveness through grant CSO2017-82881-R.
Notes
Supplemental material
Supplemental material for this article is available online.
Author biography
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
