The impact of exogenous triggers on institutional change

Early studies on electoral reforms were developed as a result the concomitance of three major electoral reforms in 1993 in Japan, Italy, and New Zealand. These three countries were struck by a period combining intense political dissatisfaction and rapid electoral change, and the literature that developed to explain the new electoral systems adopted focused on these two intertwined aspects (for a synthesis, see Gallagher and Mitchell, 2005 and Renwick, Reference Renwick2010). These findings on the role of political crisis, however, have not often been systematized in large-N studies. Second, the ‘crisis’ explanations of institutional change deal with exceptional situations rather than with everyday politics. Indeed, the context in which the 1993 electoral reforms took place was uncommon in many ways, as the Japanese, Italian, and New-Zealander systems had shown signs of malfunctioning (corruption and low governability for Japan and Italy, reversed majorities for New Zealand) for quite a few years. The following question remains unanswered: what would be the link between ‘relative’ political crisis and institutional change, that is, between periods in which the political system is contested, but not truly endangered?

Norris (Reference Norris2011) has investigated the statistical link between democratic legitimacy understood broadly and electoral reform. Comparing the countries included in the World Values Survey between 1993 and 2004, she shows that there is a relationship between what she calls ‘democratic aspirations’ (support for democratic ideals) and electoral reform. She does not find evidence of a link between confidence in institutions, evaluation of the democratic performance of a country, and electoral reform. Norris includes a great variety of countries and studies the link between the overall level of democratic legitimacy and reform (i.e. the long-term impact of this factor), rather than the impact of short-term shifts on the propensity to reform. Whereas Norris studies the level of legitimacy as an ‘inherent’ factor of reform (Shugart in Shugart and Wattenberg, Reference Shugart and Wattenberg2003), one might argue that shifts in the level of trust in political institutions and overall appreciation of the democratic performance might be ‘contingent’ factors for reform (i.e. short-term triggers).

The literature on electoral system change has progressively attempted to go beyond crisis explanations by evaluating the link between shifts of political competition, the previous state of the party system, and electoral reforms. Several authors have ‘reversed’ Duverger’s laws (Colomer, Reference Colomer2005; Benoit, Reference Benoit2007), turning around the common assumption that electoral laws shape electoral outcomes, and ultimately, party systems. Doing so, these authors have shown that, in countries such as Belgium or the Netherlands, for example, which adopted proportional representation (PR) at the beginning of the 20th century, the proliferation in the parliament of multiple parties preceded the adoption of PR. A number of authors focus on the switch from Single-member plurality to PR in Europe at the beginning of the 20th century. They show how the emergence of socialist parties as a threat to incumbent elites (Rokkan, Reference Rokkan1970; Boix, Reference Boix1999), the correction of partisan biases caused by majoritarian electoral systems after the extension of electoral franchize (Calvo, Reference Calvo2009), or contexts of extreme political uncertainty for the main party over its future electoral success (Andrews and Jackman, Reference Andrews and Jackman2005), have led to proportional electoral reforms in Western Europe.

Andrews and Jackman provide a very valuable piece of research for our purposes here. They demonstrate how uncertainty encourages short-termist behaviour of political parties, and miscalculations about the effects of new institutional rules. They argue: ‘For political actors to engage in reform of the procedures by which they won in the first place, they must come to believe either that existing arrangements will adversely affect their future prospects for winning, or that they face considerable uncertainty, or both’ (Andrews and Jackman, Reference Andrews and Jackman2005: 66). The same study has also established a clear link between uncertainty and electoral shifts. Focusing on the more recent period, Remmer (Reference Remmer2008) shows how in Latin-America over the last decades, the degree of electoral volatility as well as the variation in the effective number of political parties contribute to explain the frequency of electoral system change.

Finally, some authors have analysed a fourth factor: the impact of winning and losing elections for individual and party support for change. There have been two main interpretations of the impact alternation on the support for reform. On the one hand, some authors argue that different parties hold different preferences regarding reforms, and that some parties are, for normative or strategic reasons, more prone to support reform than others. On the other hand, other authors develop a slightly more complex interpretation arguing that incumbency had a systematic effect, leading parties that have been in power for a long time to be less reform-prone. Regarding the first interpretation, a number of accounts have also introduced a simple, but central aspect: that governing and opposition parties tend to have different preferences regarding the ideal electoral system. This conclusion is important, in that the previous accounts on electoral system change focused mainly on the self-interested motivations of political actors (Rahat, Reference Rahat2004; Pilet, Reference Pilet2008), arguing that electoral systems are redistributive in essence in that it implies winners and losers (Tsebelis, Reference Tsebelis1990). More recent works show that normative motivations about the ‘best’ system, regardless of self-interest and ideology, are central to explaining the positions of individuals and parties on electoral reforms (Bol, Reference Bol2013). Dealing with minor electoral reforms, Jacobs and Leyenaar (Reference Jacobs and Leyenaar2011) have already demonstrated with the example of the Netherlands that, when the consequences of a reform for the distribution of powers are unclear for political parties, their support or opposition to reform was more likely to be the consequence of their normative conceptions about the desirability of a particular reform.

The second interpretation of alternation, resulting mainly from studies of the individual opinions of MPs on electoral reforms, has demonstrated a consistent link between the fact of being an incumbent, being in the majority, and a lowest support for electoral reforms (Bowler et al., Reference Bowler, Donovan and Karp2006). This result confirmed previous findings dealing with the support of political elites for the adoption of direct democracy mechanisms, in which Bowler et al., Reference Bowler, Donovan and Karp2002 showed that incumbents and members of governing parties were significantly less supportive of such reforms. In a similar vein, Pilet and Bol (Reference Pilet and Bol2011) demonstrate, analyzing the positions of parties on 13 electoral reform debates, that parties that have spent more than 60% of the time in government in the last 25 years were significantly more risk-averse and more supporting of the existing electoral arrangements, no matter what the expected consequences of the reform were in terms of seats.

To summarize, theoretical and empirical links have been established between situations of political crisis, shifting support for the democratic system, rising electoral uncertainty, winner and loser status, and the propensity for electoral system change. These four factors can be thought of as four facilitating conditions for institutional reforms. I argue that these factors also apply to a wider set of reforms, including a broad range of reforms of the core democratic rules. In fact, there are strong reasons to believe that these factors should play an even stronger role for democratic reforms in general. Electoral system change is a rather pure example of redistributive reform involving zero-sum games. Elites have strong vested interests in the existing electoral systems, but the existing literature has already proven, in the circumstances examined above, that these interests and other barriers can be overcome and lead to reform. More and more authors insist on the fact that, outside of outcome-contingent motivations (based on the expectations about the expected consequences of a reform), act-contingent motivations (based on the expectations about the consequences of the very act of supporting a reform) are central in explaining the support for democratic reforms (Renwick, Reference Renwick2010, Reference Renwick2011). In the case of reforms where the consequences, in terms of distribution of power, are more uncertain than pure redistributive reforms such as electoral reforms, elites should also base their judgments on act-contingent motivations and pay strong attention to exogenous changes, such as crisis, electoral shifts, and voter satisfaction to make up their mind. In addition, the literature on electoral system change has also early on, demonstrated that not all institutional systems are equally sensitive to short-term shocks and vulnerable to change.

Endogenous factors of change: the varying vulnerability of different systems towards change

Shugart has first suggested the interaction between exogenous and endogenous factors during processes of electoral reforms, arguing that electoral reforms are the result of the conjunction between what he calls ‘contingent’ events, and ‘inherent’ characteristics (an electoral system considered as extreme for the dimensions of personalization of the votes and the concentration of power in the government, Shugart in Shugart and Wattenberg, Reference Shugart and Wattenberg2003). He has shown that both Italy and Japan, before the reforms of the 1990s, displayed extreme characteristics regarding the personalization of votes that fuelled reform in combination with corruption scandals and political crisis. Other authors, working on the early adoption of PR in Europe, have shown that, in addition with the exogenous factors such as socialist vote or electoral uncertainty, the switch to PR was more likely in countries using majority systems than in countries using plurality systems (Blais et al., Reference Blais, Dobrzynska and Indridason2004). These authors argue, indeed, that majority systems imply a more fragmented party system, more coalition governments, higher uncertainty over the electoral outcomes and strategies, which all lead to stronger incentives to adopt PR.

In this article, I investigate the impact of short-term shocks in terms of the institutional change that might differ across systems, once the difficulty to reform has been controlled for. This idea is closely linked with the theory of the determinants of policy stability, that is the veto-players approach, introduced by Immergut (Reference Immergut1990, 1992) and systematized by Tsebelis. Tsebelis’s (Reference Tsebelis2002: 2) core argument posits that ‘to change the legislative status quo, a certain number of individual or collective actors have to agree to the proposed change’. Consequently, the opportunities to change the status quo depend on the number of veto players and on the ideological distance between them. He concludes by predicting that ‘the countries with many veto players will have policy stability, while the ones with one veto player may have instability’ (Tsebelis Reference Tsebelis2002: 5). Institutions are typically defined as ‘stable, valued, recurring patterns of behaviour’ (Huntington, Reference Huntington1965). Many authors have argued that changing institutions is (even) harder than changing policies (Colomer, Reference Colomer2001: 235): ‘Standard models in political science and public policy studies basically focus on three elements: citizens’ preferences, political party’s or candidate’s positions, and institutional rules. The typical assumption is that the set of institutional rules is the most stable of these elements’. Using the case of electoral reforms in Belgium, authors such as Hooghe and Deshouwer (Reference Hooghe and Deschouwer2011) also demonstrate that the barrier constituted by veto players is very difficult to overcome most of the time. Empirically, too few longitudinal and cross-national studies on institutional reforms have been conducted to be able to accept or to reject the postulate that policy change works in the same way as institutional change. Therefore, it is necessary to control for the impact of partisan veto players (number of parties in government) and institutional veto players (here approximated by controlling for country effects) on the number of reforms.

Hypotheses

The main theoretical assumptions of the article can be summarized quite simply. Drops in the satisfaction with the democratic system, rising electoral uncertainty, and the arrival in power of new political forces with distinctive preferences encourage institutional reforms. This study enquires about three main variables: the variation in the support for the functioning of the democratic system (measured by shifts of satisfaction with democracy), the rise of electoral uncertainty (measured by shifts of volatility), and the change of preferences of the actors getting into government (measured by alternation). The two first explanatory variables translate changes in the environment in which the political parties compete, while the third one is focused on shifts of preferences of actors in power. The number of veto players, the level of constitutional rigidity, the ideological orientation of the governments in the legislature, and economic growth are used as control variables:

Expanding the assumptions of Norris, this first hypothesis posits that when support for the democratic system drops, the pressure to adopt institutional reforms is stronger, as parties have to react to public dissatisfaction with the democratic system through institutional change. This postulates, first, that a drop in the level of political support translates the perceptions of the malfunctioning of the institutional system, and of dissatisfaction of citizens about the way it works. Second, it posits that political parties are actually able to perceive these drops and to react to them:

The level of volatility can be thought as the tangible expression of shifts of electoral preferences from one election to another, and of the level of electoral uncertainty in which political actors compete. Each political system has an inherent level of volatility linked with the structure of its party system. The rise in the level of volatility compared with this ‘structural’ level, on the other hand, is the translation of rising uncertainty about the balance of power between the different parties. This in turn may well be the result of dissatisfaction with the political system, as illustrated, for example, by the unprecedented levels of volatility observed in Italy in 1994, or in Ireland in 2011, in the aftermath of a very severe crisis of legitimacy for political elites. In such a case, volatility implies a level of uncertainty affecting all parties. It might also be the consequence of dissatisfaction with the incumbents’ governing performance. In both scenarios, I hypothesize that when the level of volatility rises from one election to another, one should witness more institutional reforms in the following legislature:

Third, I expect that political alternation, as it brings to power parties that were previously part of the opposition in power, is a strong incentive for institutional reforms. Alternation provides, to some, political opportunities that they did not have during their time in opposition to carry on reforms. This assumes that governing and opposition parties have built over time different preferences upon institutional reforms. This hypothesis also implies that when previous losers come into government, they are provided with political opportunities to change a system they were less satisfied with than previous incumbents, expanding the conclusions about the fact that parties previously in opposition are more reformist and less risk-averse. Alternation is therefore understood here, primarily, as a switch of preferences of the parties composing the government.Footnote ¹

The dependent variable: the number of institutional reforms adopted by legislature

This article uses the database ‘Institutional Change in Advanced European Democracies’, collected by Alex Wilson and developed within the realm of the SIEPOL research project (Seclusion and Inclusion in the European Polity: Institutional Change and Democratic Practices) directed by Peter Mair and Adrienne Héritier. It covers over 20 years of institutional reforms (1990–2010) in 18 European countries that democratized before 1989: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Luxembourg, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, and United Kingdom.Footnote ² The data set encompasses six categories of institutional reform: electoral system reform, parliamentary reforms, federal reforms, or reforms of decentralization, direct election of the head executive at the national or local level, provisions for referendums and citizens’ initiative at the national level, and finally, the regulation of the access to suffrage. The interest of this data set is to cover a wide, although of course non exhaustive, range of reforms of the core democratic rules. Some of the institutions identified, for example, by Lijphart (Reference Lijphart1999), such as constitutional review, are unfortunately absent. Others, including reforms of direct democracy, are included. Although the aim of the SIEPOL database was to be as exhaustive as possible for the six dimensions included, some reforms may have been omitted. Despite its limitations, the sample of reforms is however the largest currently available.

It is important to note that what counts as a ‘case’ of reform is the modification of the formal rules on one of the six dimensions. This means that if one single law reforms three different dimensions at once, it is considered that three reforms took place, just as if three different laws are adopted. For example, France adopted a constitutional reform in 2008 that contained at once a parliamentary reform, the introduction of constituencies for citizens living abroad, and new provisions about the organization of popular referendums, that were included in the database as three reforms (Bedock et al., Reference Bedock, Mair and Wilson2012). In the following regressions, the unit of analysis is the legislature and the dependent variable the number of reforms adopted in a given legislature. For the models taking into account all the reforms, this number ranges from zero to 7 (Table 1), with almost two-thirds (63%) of the legislatures comporting at least one institutional reform. The number of reforms adopted was preferred to a different dependent variable such as a dummy variable reform/no reform. Statistically, it offers a much more sensitive measure, without assuming arbitrary thresholds, considering, for example, that the crucial difference to investigate is between zero reform and some reforms. The literature presented above has mainly focused on the occurrence of reform rather than on its frequency. Substantively, however, it makes sense to focus on the number of reforms insofar as the database at our disposal accounts for many more reforms than the existing data sets. Indeed, this data set enables the evaluation not only the impact of a given variable on reform vs. no reform, but also its impact on the number of reforms more generally.Footnote ³ Moreover, the choice to aggregate the adopted reforms avoids the composition of the different dimensions of reform from affecting the statistical results, since each reform is taken into account.

Table 1 Number of institutional reforms adopted by legislature in 18 Western European democracies (1990–2010)

Source: My own elaboration of the SIEPOL database (Bedock et al., Reference Bedock, Mair and Wilson2012). Compared with the previous version of the SIEPOL database, two additional reforms that happened in Ireland were included in the analysis: the reform of the constitutional amendment to recognize the role of local government adopted in 1999 and the 2001 Local Government Act.

The first question that arises is to find the appropriate method for investigating this dependent variable, characterized by several aspects: its distribution is not normal or discrete, and it constitutes count data.

Specification of the model

The analysis has been conducted using negative binomial regression, a subcategory of event count models. Event count models are statistical models in which the dependent variable is a count of events, here, the number of institutional reforms adopted in a given legislature, therefore consisting of discrete, nonnegative integer numbers. They constitute a generalization of the linear model. The traditional OLS regression cannot account for the discreteness and functional form of the dependent variable, which would lead to biased estimators. These models are estimated using maximum likelihood.

To build adequate statistical models based on this dependent variable, the negative binomial regression was chosen over the Poisson regression, because of the issue of over-dispersion of the data (see Appendix 3). Indeed, Poisson regressions have stringent requirements, that is, that the conditional mean is equal to the conditional variance. The existence of over-dispersion can be interpreted as the violation of the requirement that counts are independent from one another (Hilbe, Reference Hilbe2011: 2). Its consequences on the estimators can be interpreted in a similar way as the violation of the assumption of homoskedasticity in the linear regression model, meaning that the goodness of fit of the model is overestimated (Cameron and Trivedi, Reference Cameron and Trivedi2001). The overall fit of the models was estimated using the likelihood ratio test, which compares the fit of the estimated model with a null model where all parameters are held at zero.

Explanatory variables and control variables

The first explanatory variable concerns the shift in the satisfaction with the way democracy functions. Dalton has distinguished between no less than five objects of political support: confidence in public authorities (parties, politicians), confidence in political institutions (government, parliament), the evaluations of the regime’s performance (satisfaction with the functioning of the democratic system), support for democratic principles, and finally, support for the political community. While the shifts of the first aspect express ‘no more than dissatisfaction with the incumbents in office, a normal and healthy aspect in democratic process’ (Dalton, Reference Dalton1999: 57), and the two last aspects, the support for the polity itself – which can be considered as unproblematic in consolidated Western democracies – the second and the third dimensions of political support are interesting in the perspective I follow. Ideally, testing for the impact of shifting support for political institutions, and satisfaction with the way democracy is working in a given country, could provide the appropriate answers to the questions of the link between shift of support and institutional reforms. However, assessing appropriately the shifts between one legislature and another would require a large number of data points in time, ideally each year. Such data only exists for satisfaction with democracy, meaning that I will restrict the analysis to this aspect of political support. The variable used is the shift in percentage of the aggregate level of respondents in a given country who say they are ‘very satisfied’ or ‘fairly satisfied’ to the trend question of the Eurobarometer ‘are you satisfied with the way democracy is functioning in your country?’.

The second explanatory variable is the shift in percentage in the level of total volatility, defined by Bartolini and Mair (Reference Bartolini and Mair1990: 17) as the ‘measure of the net electoral change between two consecutive elections’, and using the Pedersen (Reference Pedersen1979) index. This second explanatory variable is a proxy for the level of uncertainty of each election, with the underlying idea that the more important the shifts between one election and the other, the more unstable the party system and the electoral environment.

The third and final explanatory variable is political alternation at the beginning of a given legislature. This is a dummy variable defined as the concomitance of a change in the political orientation of the cabinet and a change in the head executive. The ideological orientation of the cabinets in each legislature is the result of a specific literature review to correctly classify parties in each country (see online Appendix 2).

The first control variable is the aggregated number of partisan veto players in the government for each legislature (measured as the number of parties present in the government), therefore controlling for eventual effects of the government configuration on the occurrence of institutional reforms. This measure of partisan veto players has the advantage of being very simple, and as reliable as the more complex measures that have been tested in the literature, as shown by Roller (2003, Reference Roller2005). Since this data is composed of legislatures by country, I also control for country effects by including 18 dummy variables for each of the countries included. The third control is the ideological orientation of the cabinets in the legislature: left/centre left, right/centre right, and grand coalition. Given the development of the reflection about the link between economic crises, political legitimacy and institutional reforms, I also included the average level of economic growth in each legislature as a control variable, although the timeframe of this paper stopping in 2010 does not really enable me to conduct precise investigations on this point.Footnote ⁴