Is there a linear relation between institutional constraints and reform of democratic rules?

5In this article I focus on a series of democratic reforms that are the object of legislation at national level. Institutions are typically characterised as “stable, valued, recurring patterns of behavior”. [14] In order to modify democratic institutions a certain number of barriers must be overcome which are variable from one political system to another. These barriers take the form of what Roger D. Congleton calls “laws for making laws”, [15] or of the existence of a certain number of actors capable of vetoing a reform.

6In a foundational study on the subject, Donald S. Lutz demonstrated that there is an empirical interactive link between the duration of a constitution, the difficulty amending it, and the frequency of modifications to it. [16] Building on Lutz’s results, other authors have shown the importance of what they call the “cost effect” to explain the frequency of constitutional reforms. Bjorn Erik Rasch and Roger D. Congleton thus show how “amendment procedures may contribute to both the stability and durability of a constitutional regime”. [17] Other studies, which have explored the question of the durability of constitutions, have demonstrated that there is a positive, but non-linear, relationship between constitutional rigidity and how long constitutions are in place. In effect, even though these two elements are linked, a constitution that is too rigid is not open to adaptation, which can lead to the complete replacement of one constitution by another. [18] Thus, all these studies establish a connection between the difficulty of reforming institutions, the frequency of constitutional amendments, and how long constitutions last. If we were to apply these conclusions to the reforms under investigation here, we might anticipate that there would be a lower frequency of reforms when the procedure for constitutional reform was most restrictive.

7The idea of a relationship between institutional constraints and stability is the principle argument of the research that introduced and developed Ellen M. Immergut’s notion of the “veto point”, [19] systematised by an approach in terms of George Tsebelis’s concept of “veto players”. [20] Tsebelis’s main argument is that “to change the (legislative) status quo[,] a certain number of individual or collective actors have to agree to the proposed change”. [21] As a result, the capacity to alter the status quo depends on the number of veto players and the ideological distance separating them. He concludes by predicting the stability of public policies in countries characterised by a large number of veto players, whilst countries with a single veto player are likely to experience greater instability. In more recent versions of his argument, Tsebelis has limited himself to claiming that “One analytical truth connecting veto players with policy stability is that as the number of v[eto] p[layers] increases, policy stability does not decrease”. [22] If his argument applies beyond public policies, to democratic reforms, we should observe many more reforms in countries generally governed by a single party, such as the United Kingdom, and less reform in countries governed by large, ideologically heterogeneous coalitions, such as Belgium. We can also see the underlying idea of a connection between veto players and the stability of pre-existing formal institutional rules in the work of James Mahoney and Kathleen Thelen, although these authors propose an approach to institutional change that is quite different from that of Tsebelis. They predict for example that “displacement” change – in which existing rules are replaced by new rules – is more likely when actors’ opportunities to exercise a veto are few and where the margin for interpreting the rules is reduced. In circumstances where vetoes are highly likely, other types of changes to the rules are more probable: actors will tend to “drift” (re-interpreting existing rules) or to “layer” (adding new rules to existing rules). [23]

8Other authors have developed a less linear vision of the connection between veto players and stability. Unlike Tsebelis, Markus M. L. Crepaz distinguishes the impact of collective veto players and partisan veto players. Where the presence of several collective veto players – such as several parliamentary chambers for example – might encourage the adoption of reform through a process called “logrolling”, the presence of several partisan veto players may lead to inertia in redistribution policies. [24] Other authors, such as Gideon Rahat, who have specifically looked at structural reforms in democratic regimes, consider that there is a curvilinear relationship between the number of veto players and the number of reforms. [25] According to this perspective, reforms are easier when power is either very concentrated, or when it is so dispersed that no single actor can block reform alone. This idea has also been put forward by Joel S. Hellman who has demonstrated that, in the case of Eastern Europe, a high number of veto players may also foster a greater capacity for change, in situations where it is more difficult for each party to oppose the reforms, thanks to large and inclusive coalitions. [26] However, it is generally expected that a positive linear relationship exists between the degree of institutional constraint and the difficulty of introducing and implementing reform.

9In order to evaluate the relevance of the institutional approach, this article will examine the hypothesis that the degree of institutional constraint, measured by constitutional rigidity and the number of veto players, is a hindrance to reform; bearing in mind the reservations of several authors who have suggested that this relationship is not necessarily either linear or absolute.

The relationship between the level of political support and the reform of democratic rules

10Russell J. Dalton has described the “erosion of political support” that has affected all advanced industrial democracies and Western Europe in particular. [27] Political parties and democratic institutions are confronted with unprecedented challenges to their legitimacy. For Dalton, the level of popular support can be broken down into five distinct objects: confidence in public authorities (parties and politicians), confidence in political institutions (government, parliament), evaluations of regime performance (the level of satisfaction regarding the functioning of democracy), the level of support for democratic principles, and the level of support for the political community. Dalton observes that “public dissatisfaction now reaches beyond the political authorities to the institutions and norms of the democratic process”. [28] He adds that “as the object of dissatisfaction becomes more general – the regime, the political community – the political implications broaden. A decline in support for the political process might provoke a basic challenge to constitutional structures or calls for reform of the procedures of government”. [29] In a collected work on the transformations of contemporary democracy edited by Cain, Dalton and Scarrow, the hypothesis connecting the decline of political support to an increase in institutional reforms seems to be confirmed. The authors demonstrate an increase in institutional activity which can be seen, for example, in the multiplication of electoral arenas in which citizens can express themselves, the rise of decentralisation reforms, the development of forms of direct candidate selection, an increase in the number of measures to improve transparency, etc. [30] In other words, the central argument of the book is that “the cure for democracy is more democracy”. [31] The result is a series of institutional reforms designed to encourage citizens’ participation in the political process.

11Pippa Norris was one of the first to try and establish a statistical connection between democratic legitimacy and electoral reform. By comparing the countries included in the World Values Survey between 1993 and 2004, she demonstrated that there is a positive connection between “democratic aspirations” (the support for the democratic ideal) and the subsequent use of electoral reforms. She did not, however, find any significant relationship between confidence in institutions, evaluation of democratic performance, and the use of electoral reforms. [32] This research established several important observations. The first is the impact of public opinion and the level of political support on the likelihood of reform. The second, which stems directly from the first, calls into question the idea that a process of institutional reform results purely from interactions between political elites in which citizens are merely incidental. Lack of political trust is interpreted here as an inherent factor in reform that intensifies the question of electoral reform and encourages the implementation of revision.

12In summary, whether they are interested in democratic institutions overall or in electoral reform in particular, these authors suppose a negative relationship between the level of political support and the use of reforms, given that those democracies where this support is the weakest have a greater tendency towards democratic reform.

Electoral uncertainty and democratic rules

13The level of political support may be considered as an element that structures the environment in which political parties evolve. Yet if we look more specifically at the moment the reforms take place, the question of short-term evolutions occurring particularly in the electoral arena becomes central for explaining the frequency of these reforms. The literature on the changes to electoral systems has established a link between electoral uncertainty and electoral reform, demonstrating that the parties’ varying electoral fortunes have a clear impact on their propensity to support change.

14Authors like Josephine T. Andrews and Robert W. Jackman have focused on the impact of uncertainty on the behaviour of political actors who are able to implement reforms, ironically called “strategic fools”. [33] They show how, in a context of extreme uncertainty, political parties are encouraged to act according to short-term rationales which can lead to errors of judgement regarding the anticipated effect of new rules. Their central argument is that “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”. [34] Their study shows a clear link between uncertainty and electoral change, demonstrating that parties base their support for the shift from first-past-the-post to proportional representation on their own performance in the most recent election. Thus the parties that had performed well tended to argue for the conservation of the existing electoral system, whereas those which had not done so well promoted the introduction of proportional representation.

15Other studies have underlined the impact of volatility on electoral reforms. Once again, the increase in volatility is interpreted as an indicator of the level of uncertainty that the reformers (here essentially the parties) must face. Karen L. Remmer thus shows how in Latin America in recent decades, electoral reforms have been the result of electoral volatility, whilst simultaneously feeding this volatility via the re-adjustments introduced by the new rules. [35] Finally, a very recent study on 31 European countries between 1945 and 2012, which came out of the “Electoral System Change in Europe since 1945” project which covers multiple aspects of the electoral system (electoral formula, but also magnitude, size of assembly, type of vote, etc.), has established not only that an increase in volatility increases the probability of reforming the electoral system, but that the arrival of new players in the electoral arena pushes parties to adopt reforms that make the system less inclusive. [36] In this perspective, the electoral reforms brought about by electoral volatility represent a direct response to the electoral threat posed by new parties, as Carles Boix had already shown in his study of the impact of the rise of socialist parties on the shift from a majority voting system to proportional representation in Western Europe at the beginning of the twentieth century. [37]

Political alternation and the reform of democratic rules

16In attempting to evaluate the impact of electoral evolutions on democratic reform, a number of authors, often in the context of the study of electoral reform, have demonstrated the impact of electoral success or failure on a particular party’s support for change. The impact of political alternation has been interpreted in two ways. On one hand, certain authors have developed the idea that different parties have different preferences regarding reforms, some being quicker to support reform than others for normative or strategic reasons. Other authors, however, have adopted a slightly more complex interpretation, revealing the impact of the status of the incumbent party: parties that have spent a long time in power are less likely to be favourable to institutional change.

17The first interpretation thus argues that parties in opposition and those in government tend to have different preferences as to the ideal electoral system. This idea is important given that it is now well-established that actors’ selfish motivations are central to understanding their position on electoral reform, [38] to the extent that they are essentially “redistributive” and necessarily involve the designation of winners and losers. [39] Other more recent work has also emphasised the fact that the positions of individuals and parties regarding electoral reform are largely dependent on their normative preferences regarding what they consider to be the best system, independent of ideology or the maximisation of interests. [40] Focusing more specifically on minor electoral reforms, authors like Kristof Jacobs and Monique Leyenaar have shown that, in the case of the Netherlands, when the consequences of a particular reform on the distribution of power are not clear for the political parties, their support of or opposition to this reform depends essentially on normative considerations as to the desirability of the reform. [41]

18The second interpretation of the role of alternation, which is primarily based on the study of parliamentarians’ opinions of electoral reforms, establishes a clear link between being an outgoing MP, or belonging to the majority, and a low probability of supporting electoral reform. [42] This result confirms similar conclusions regarding the support of political elites for adopting mechanisms of direct democracy. [43] From a similar perspective Damien Bol and Jean-Benoît Pilet have analysed the position of parties in thirteen electoral reform processes, revealing that the parties which had been in government more than 60% of the time over the previous 25 years were significantly more likely to support the existing electoral system, whatever the consequences to be expected in terms of seats. [44]

19However, most empirical investigations have concentrated on electoral reforms rather than democratic reforms more generally. But can we expect there to be similar effects on democratic reforms in general of institutional constraints, political support, volatility, and alternation? There are reasons to think that these factors may play out even more strongly in relation to broader democratic reforms. For authors such as Josep Colomer, changing institutions is more difficult than changing public policy. He states that “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”. [45] Institutional constraints should therefore play an important role. Second, electoral reforms are relatively “pure” examples of zero-sum redistributive reforms, in which elites have powerful personal interests. However, it has already been demonstrated that these personal interests and other barriers can be overcome and lead to reforms. Authors taking this view increasingly stress that, beyond those motivations that are contingent on consequences of the reforms and based on the anticipation of their results (outcome-contingent motivations), the motivations based on the act of reform (act-contingent motivations) are also central to explaining support for electoral reform. [46] In the case of reforms where the consequences in terms of the distribution of power are more uncertain than for electoral reforms, there are strong reasons to think that the reformers base their judgments on motivations linked to the perception of the act of reform itself and are more sensitive to changes in political opinion, whether manifested in the level of public support or via the ballot box.

Distinct long-term and short-term factors

20On the basis of the theoretical considerations outlined above, four main hypotheses are developed here in two distinct empirical models. This article aims to show that it is necessary to distinguish clearly the factors that create favourable conditions for change in democratic institutions in the long term, as well as the short-term shocks that explain why a greater number of reforms occur at a particular time.

21H1. In the short term and the long term, the greater the degree of institutional constraint, the less a country reforms its institutions.

22Here I seek to test the relevance of institutional logic in order to establish whether there is indeed a negative linear relationship between the degree of institutional constraint and the frequency of reform in the long term and the short term. In other words, this hypothesis seeks to verify whether the rationale proposed by Tsebelis or Lutz, that a greater number of veto players and more rigid institutions hamper reforms, is in fact relevant in analysing the frequency of democratic reforms.

23H2. In the long term, the lower the level of political support for institutions and for the democratic system in a country, the more likely it is to reform its institutions.

24We would expect that elites in power tend not to reform democratic institutions unless they feel constrained to do so because of low levels of political support, and thus of the dissatisfaction of citizens towards their political system because of its lack of legitimacy. This hypothesis also supposes that political elites in a position to reform institutions consider that a minimum level of support is necessary and desirable for democratic institutions to function. An alternative hypothesis, which will be tested, might consist in thinking that elites react to the level of political support not only in the long term, but also in the short term (and thus that they perceive and react to a drop in the level of citizens’ confidence regarding the functioning of their democracy and its institutions).

25H3. In the short term, when the level of volatility in the electoral arena increases, the number of institutional reforms adopted in the subsequent parliamentary term increases.

26The level of electoral volatility can be perceived as the tangible expression of variations in electoral preferences from one election to another, but also as an indicator of the level of electoral uncertainty that political actors face. Each political system is characterised by an inherent level of volatility that is linked to the structure of its party system and the number of parties in competition. Any increase in this volatility can be analysed as an increase in the uncertainty regarding the balance of power between the parties. Such an increase in uncertainty might also be the result of increased dissatisfaction with the political system, such as for example the record levels of volatility recorded in Italy in 1994, or in Ireland in 2011, following a severe crisis of legitimacy for the political parties. In such circumstances the increase in volatility implies a high level of uncertainty affecting all political parties. In either scenario, we would expect that an increase in the level of volatility means that we can predict that more reforms will be adopted during the subsequent parliamentary term. In order to test that the impact of this variable is limited to the short term I will also test the impact of the average level of volatility on the number of reforms in the long term.

27H4. When actors formerly in opposition gain power, the number of institutional reforms adopted in the subsequent parliamentary term will increase.

28By definition, political alternation means the arrival of the former opposition in power. Here we would expect that alternation constitutes a strong incentive to adopt institutional reforms, in that it gives new parties the political opportunity to bring about reform, a capacity that they did not previously possess. This article therefore works from the assumption that the governing parties and the opposition parties have developed, over time, different preferences regarding democratic reforms. We might also suppose that when former opposition parties arrive in government they have the political capacity to reform a system that they were less satisfied with than the outgoing party. This proposition is connected to the idea that opposition parties tend to be more reformist and less inclined to risk-aversion. In any case, it is presumed that alternation brings parties with different institutional preferences to power and therefore that a greater number of reforms occur in a given parliamentary term when the incumbents lose the elections.

A longitudinal and multidimensional analysis of democratic reform

29To test the four previous hypotheses and the impact of the degree of institutional constraint, the level of political support, the increase in volatility, and political alternation on the frequency of institutional reform, this article uses a database called “Institutional Change in Advanced European Democracies”. This database was developed in the context of a research project called SIEPOL (Seclusion and Inclusion in the European Polity: Institutional Change and Democratic Practices) [47] and it covers 20 years of institutional reform (1990-2010) in 18 European countries that transitioned to democracy before 1989: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Luxemburg, Norway, the Netherlands, Portugal, Spain, Sweden, Switzerland, and the United Kingdom. [48] It covers six categories of reform: electoral reform, parliamentary reform, federalisation and decentralisation reform, direct election of the head of the executive at the national or local level, regulations concerning referenda and citizens’ initiatives at the national level, and finally the regulation of the right to vote. This database has the advantage of covering a large (although not exhaustive) range of reforms to basic democratic rules. It is true that some of the institutions identified by Lijphart, such as the regulation of constitutional justice, are missing. [49] Others, such as the question of direct democracy, are included. However, this sample of reforms remains one of the largest available and enables statistical analyses concerning a number of established democracies.

30It is worth noting that the modification of the rules on each one of these six dimensions is counted as a case of reform. In other words, if a single law modifies x number of dimensions in the institutional architecture, then we consider there to have been x number of reforms. This was for example the case during the French constitutional reform in 2008, which included parliamentary reform, the introduction of constituencies for French overseas voters, and new rules for the organisation of popular referenda.

The determinants of reform in the long term: description of variables

31The first series of empirical investigations, which aims to test the first two hypotheses, considers the 18 democracies included in the database as units of analysis. The dependent variable used is the number of reforms adopted per country between 1990 and 2010, ranging from 3 (Denmark) to 17 (France). The analysis consists primarily of a series of descriptive statistics (scatter plots, and simple linear regressions) and an analysis of the correlation coefficients. The small number of observations makes more elaborate analytic techniques such as multiple linear regression impossible.

32The main variables analysed concern first the structural institutional constraints of each political system, to verify whether the systems in which there are stricter rules governing institutional reform and where there are a higher number of veto players in government reform less than the others. I have therefore examined the average number of partisan veto players over the period, measured by the average number of parties in government between 1990 and 2010 in each country. The second institutional variable concerns the level of constitutional rigidity measured on a scale from 1 to 4, from the least to the most rigid. [50] The distinction between what Tsebelis calls institutional veto players and partisan veto players is justified here by research conducted with synthetic indicators which shows that that the indicators which combine structural institutional characteristics and competition characteristics tend to measure different concepts. [51] The measures selected here are as parsimonious as possible. The decision to focus on constitutional rigidity is justified by the fact that I am interested here in the obstacles that the constitution throws into the path of reformers. Regarding partisan veto players, Edeltraud Roller has demonstrated that the simplest measure (the number of parties in government, unweighted) is just as powerful as more complex measures. [52]

33One of the variables that are of the most interest to us here is the measure of the average level of political support recorded in each country between 1990 and 2010. I have constructed a scale of political support that theoretically extends between 0 and 100%, combining a number of measures: the level of confidence in political parties, in the parliament, and in the government, and the level of satisfaction with the functioning of democracy in the country. The scale takes the average of respondents who say they are “fairly” or “very” confident in the above institutions, and those who are “fairly” or “very” satisfied with the functioning of democracy, using the Eurobarometer data for the countries belonging to the European Union and the World and European Values Survey for Iceland, Norway and Switzerland. This level of support varies between 28.6% for Italy and 63.2% for Denmark.

34Finally, to test the impact of volatility and alternation in the long term, we look at the average level of total volatility recorded for the period between 1990 and 2010. This is defined by Stefano Bartolini and Peter Mair as the measure of net electoral change between two consecutive elections, [53] measured by the Pedersen index. [54] I am also interested in a possible long-term link between the proportion of elections over the 1990-2010 period that ended in political alternation, and the number of reforms. The countries included in this database vary significantly in terms of population. The graphs below reflect the size of the population based on the number of registered voters at the last election.

Table 1

Analysis of the long-term determinants of democratic reform in Western Europe, 1990-2010

Constitutional rigidity Average number of parties in government Volatility Proportion of political alternation Level of political support Number of reforms Germany 3 2.6 8.4 28.6 37.6 6 Austria 2 2 10.7 25 50.3 10 Belgium 3 4.8 11.7 43 41.7 15 Denmark 3 2.3 11.3 28.6 63.2 3 Spain 3 1 8.8 33.3 43.4 3 Finland 2 4 9.5 66.7 52.9 11 France 3 2.4 13.8 80 34.1 17 Greece 2 1 5.9 50 37.6 9 Ireland 3 2.3 9.6 0 43.5 6 Iceland 4 2.1 12.8 28.6 42.5 7 Italy 2 4.9 15.4 85.8 28.6 12 Luxemburg 2 2 8.5 40 60.7 5 Norway 3 2.2 15.9 66.7 59 4 Netherlands 4 2.7 17 57.1 55.6 4 Portugal 2 1.1 14.9 42.9 38 13 United 1 1 7.8 33.3 38.1 12 Kingdom Sweden 2 2 13.4 50 50 6 Switzerland 2 3.9 7.9 0 53 4

Analysis of the long-term determinants of democratic reform in Western Europe, 1990-2010

The frequency of reform in the long term: the importance of the level of political support

35If we look at the correlation coefficients between the number of reforms and, respectively, the level of constitutional rigidity and the number of parties in government, we can see that there is a slightly negative relationship between constitutional rigidity and the number of reforms, and a slightly positive relationship between the average number of parties in government and the number of reforms. However, neither of these relationships is statistically significant, as we can see in the scatter plot in Figure 1. Here we can see that a country like Belgium, characterised by a high level of constitutional rigidity and an average of almost five parties in government, has adopted a large number of reforms (15), whilst other countries with more flexible constitutions and less numerous governments have made fewer reforms. We can also see that the vast majority of countries do not fall within the 95% confidence interval (CI). This first result is important because it suggests that, in the long term and contrary to the prediction of our first hypothesis, the degree of institutional constraint, whether measured by constitutional rigidity or by the number of parties in government, is not a hindrance to reform.

Figure 1

Total number of reforms of core democratic rules against the average number of parties in government

Total number of reforms of core democratic rules against the average number of parties in government

36If we now analyse the relation between the number of reforms adopted and the level of political support, we can see, as expected, a statistically significant negative relationship between the two variables (correlation -.066, p<0.01). The scatter plot (Figure 2) of the relation between the number of reforms and the level of political support weighted by voter population in each country clearly represents a strong negative linear relationship between the two variables. The higher the level of political support (and thus high confidence in institutions and satisfaction with democracy), the fewer the number of democratic reforms adopted over the period 1990-2010. Thus, in Italy, where the level of political support was the lowest, (28.1%), the regression line predicted the adoption of thirteen reforms, whilst in Denmark, where the level of political support was the highest (63.2%), it suggests the adoption of two reforms (in reality there were three). The relationship is of course imperfect; we can see that some countries are outside the 95% confidence interval, such as France where the number of reforms adopted was higher than that predicted by the regression, or Germany and Spain where the number was lower than expected.

Figure 2

Total number of reforms of core democratic rules against the level of political support

Total number of reforms of core democratic rules against the level of political support

Note: regression coefficient -0.289 p<0.01 ***.

37Thus, these descriptive analyses seem clearly to confirm the second hypothesis, suggesting that those countries in which the level of political support is lowest have a greater tendency to reform their democratic rules than the others.

38As expected, there is no statistically significant relationship between the level of structural volatility over the period and the number of reforms adopted. On the other hand, we can see a slightly statistically significant positive relationship between the proportion (as a percentage) of legislatures that experienced political alternation and the number of reforms adopted (correlation of 0.44, p<0.1). Thus, where none of the parliamentary terms of the period between 1990 and 2010 ended in an alternation of power, the number of reforms expected was 4, whereas if all the parliamentary terms ended in political alternation the regression line suggests the adoption of close to 12 reforms.

Figure 3

Total number of reforms of core democratic rules against the percentage of legislatures experiencing alternation

Total number of reforms of core democratic rules against the percentage of legislatures experiencing alternation

Note: regression coefficient 0.082, p<0.1*

The determinants of reforms in the short term: variables and models

39To test the third and fourth hypotheses, the unit of analysis used was that of the parliamentary term and the dependent variable was the number of reforms adopted during a given parliamentary term. This number varies between 0 and 7, with nearly two thirds (63%) of parliamentary terms during which at least one institutional reform was adopted. I chose this variable rather than, for example, a dummy variable indicating only if a reform was adopted or not, so as to have a measure that it is more sensitive and less influenced by possibly arbitrary threshold effects. The first methodological difficulty stems from the fact that this is a discrete dependent variable based on count data, with a non-normal distribution.

Table 2

Number of reforms adopted per parliamentary term, 1990-2010

Reforms per parliamentary term N 0 43 37.1 1 37 31.9 2 18 16.4 3 9 6.9 4 3 2.6 5 2 1.7 6 3 2.6 7 1 0.9 Total 116 100%

Number of reforms adopted per parliamentary term, 1990-2010

40To verify the possible existence of a short-term impact of the variation in political support, ideally we would reproduce the scale developed above. However, only the variable regarding satisfaction with the functioning of democracy allows for an analysis at the level of the parliamentary term. Therefore, I have included in the following models the evolution (as a percentage) of the aggregate level of respondents saying they are “very” or “quite” satisfied with the functioning of their democracy. The second important explanatory variable is the evolution (as a percentage) of the total level of volatility from one election to another. The third and final important explanatory variable is the existence or not of political alternation at the beginning of the parliamentary term, defined as the simultaneous existence of a change in the political orientation of the government [55] and a change at the head of the executive.

41Finally, control variables are also included; first, the number of veto players in government, this time per legislature. I also control for the level of constitutional rigidity. The third control variable is the ideological orientation of the governments in power (divided into three categories: left and centre-left, right and centre-right, broad coalition). Finally, to control for the possible impact of the economic situation, I have also included in the model the average level of economic growth recorded for the parliamentary term.

42This second part of the analysis is conducted using the method of negative binomial regression, which is a sub-category of count data models. These models are used when the dependent variable counts the number of occurrences of an event over a given period and consists therefore in a discrete variable composed of positive whole integers with a high number of zeros and small values. These models allow us to generalise traditional linear models, which, if not corrected, would lead to biased estimations, because of their non-normal distribution. I have chosen to use negative binomial regression over Poisson regression because of problems encountered here with the over-dispersion of data. [56] Second, the models presented below correct standard error through the cluster method, in order to control for the problems of errors that are not independently distributed. Here we observe that the observations within each group (the eighteen countries making up the clusters) are correlated, which would pose the problem of biased standard error and thus inference if not corrected. Moreover, correction using the cluster method is preferred to robust standard errors, which normally require at least 50 groups in order to be reliable. The quality of estimations is evaluated using the Wald test, a similarity model, which compares the quality of adjustments of a given statistical model with a model in which all parameters are at zero.

The frequency of reform in the short term: the importance of upheavals in the electoral arena

43The results are reported based on the incidence rate, for reasons of interpretation, [57] with the 95% confidence interval to evaluate coefficient uncertainty.

44The first model only includes the control variables. It shows that constitutional rigidity has a negative and statistically significant impact on the number of reforms adopted in a parliamentary term, whilst the number of veto players present in a government has, all other things being equal, a statistically significant positive effect on the number of reforms adopted. Economic growth and political orientation of a government do not appear to have a statistically significant impact on the frequency of reforms. Although the relationship between constitutional rigidity and the number of reforms is in the predicted direction (greater constraints limiting the number of reforms in the short term), the positive connection between the number of parties in government and reforms adopted is counter-intuitive and is once again in contradiction with the first hypothesis. The link between institutional constraint and the frequency of democratic reforms seems therefore to obey much more complex logics than a simple negative linear relationship.

45The second model tests the effect of the evolution of the average level of satisfaction with democracy on the number of reforms adopted per parliamentary term. As suggested in the second hypothesis, the level of political support has a long-term, but no short-term, effect on the frequency of political reform. This is because we can see no statistically significant relationship between the evolution of satisfaction regarding democracy and the number of reforms adopted.

46The third model shows that, as predicted in hypothesis 3, the rise of volatility has a positive and statistically significant impact on the frequency of reforms during a given parliamentary term: one additional percentage point of volatility is associated with an increase of 3% in the number of reforms. To make interpreting the results easier, I have represented graphically the predicted number of reforms according to the evolution of the level of volatility (see Figure 4). We can see that in cases where volatility decreases by 80% from one election to another, the model predicts 0.9 reforms will be adopted, whereas if this level is multiplied by three (+200%), the number increases to 2.2. Beyond an increase of 200%, the confidence interval is much greater, which means that the relationship between volatility and the number of reforms is less clear, because there are an insufficient number of observations.

Figure 4

Predicted number of reforms according to the evolution of total volatility (model 3)

Predicted number of reforms according to the evolution of total volatility (model 3)

Note: the values of the evolution of volatility range from a drop of 77% to an increase of 357%.

47The fourth model looks at the effect of the alternation of power, and confirms the fourth and final hypothesis. Indeed, when all other variables are held constant at their mean, the predicted number of reforms is 0.9 in the absence of alternation at the beginning of the parliamentary term, and 1.7 if alternation occurs – nearly twice as much.

48The fifth model includes both the variable for the evolution of volatility and that of political alternation. Once again, the effect of the two variables is statistically significant, whereas, as in the previous model, the positive effect of the number of veto players in government on the number of reforms has disappeared. Finally the last model, which includes all the variables analysed, once again confirms the effect of the increase in volatility and alternation, with a predicted number of reforms close to those which have already taken place. Additional investigations not discussed here also demonstrate that the effect of alternation is not conditional on the level of volatility, or, conversely, that the effect of volatility is not conditional on the presence or absence of alternation of power. The models appear to converge to show that the increase in volatility and the arrival of new actors in power have an independent positive effect on the number of reforms adopted in a given parliamentary term.

Table 3

Determinants of the number of democratic reforms adopted per legislature in Western Europe 1990-2010

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Evolution of level of satisfaction regarding democracy (%) 1.006 1.005 (0.99-1.02) (0.99-1.01) Evolution of total volatility (%) 1.003** 1.002* 1.003** (1.00-1.01) (1.00-1.01) (1.00-1.01) Alternation in government 1.876*** 1.708*** 1.567*** (1.46-2.41) (1.34-2.18) (1.14-2.15) Degree of constitutional rigidity 0.753*** 0.762** 0.754** 0.776** 0.774* 0.787* (0.60-0.94) (0.60-0.98) (0.58-0.97) (0.61-0.99) (0.60-1.00) (0.60-1.04) Number of parties in 1.138** 1.103* 1.126** 1.079 1.081 1.056 government (1.01-1.28) (1.00-1.22) (1.01-1.25) (0.64-1.22) (0.96-1.21) (0.96-1.17) Ideological orientation (reference category: left/centre-left) Right/centre right government 1.221 1.221 1.232 1.310 1.308 1.281 (0.76-1.98) (0.74-2.00) (0.76-1.99) (0.82-2.09) (0.82-2.09) (0.79-2.07) Broad coalition 0.838 0.820 0.862 0.962 0.966 0.930 (0.48-1.45) (0.45-1.49) (0.51-1.46) (0.58-1.59) (0.58-1.60) (0.52-1.65) Growth 1.074 1.053 1.083* 1.083 1.088* 1.073 (0.98-1.16) (0.95-1.17) (0.99-1.18) (0.98-1.19) (0.99-1.19) (0.97-1.19) Observations 116 110 116 116 116 110 Number of clusters 18 18 18 18 18 18 Alpha 0.468 0.440 0.394 0.352 0.352 0.306 (0.25-0.86) (0.24-0.80) (0.19-0.84) (0.15-0.82) (0.13-0.84) (0.12-0.78) McFadden’s Pseudo R2 0.02 0.03 0.03 0.05 0.05 0.05 Cox-Snell’s Pseudo R2 0.06 0.07 0.10 0.13 0.15 0.15 Wald’s Chi2 17.63 20.44 14.18 30.08 31.89 42.53 Prob>Chi2 0.004 0.002 0.027 0.000 0.000 0.000

Determinants of the number of democratic reforms adopted per legislature in Western Europe 1990-2010

Note: The coefficients are reported as incident rate ratios. The 95% confidence intervals are in brackets. The dependent variable is the total number of reforms adopted per legislature between 1990 and 2010. *p<0.1, **p<0.05, ***p<0.001.

49Essentially these models allow us to draw several conclusions. First, it is clear that the institutional logic that seeks to explain the frequency of democratic reforms in terms of the degree of institutional constraint that weighs on the actors must be tempered. Indeed, although constitutional rigidity does seem to have a negative impact on the number of reforms adopted per parliamentary term (when other variables are constant), we can nevertheless see that, in the short term, it seems that the presence of a large number of parties in government is not a hindrance to the adoption of a large number of reforms; instead it may facilitate this. That reforms are governed by political logic is, on the other hand, clearly confirmed by the empirical models. We can see here that the elites are very sensitive to the events unfolding in the electoral arena when they decide to push institutional reform. An increase in volatility can be considered to be a tangible manifestation of electoral change, which results in an increase in uncertainty for political parties. By bringing parties with different preferences regarding democratic reform into power, political alternation opens a window of opportunity for these parties to implement reforms.

50* * *

51This article has sought to show that the logic which governs democratic reforms is above all political. The institutional logic that seeks to explain the frequency of reforms primarily in terms of the degree of institutional constraint present in each system is shown here to have very limited explanatory value. Although greater constitutional rigidity tends to limit the number of reforms in the short term, it has no impact on the number of reforms adopted in the long term. Even more surprising, the presence of multiple parties in government does not appear to be a hindrance to reforms, either in the short or the long term. The results presented here have thus established that there is no linear relationship between institutional constraint and the frequency of reforms. This result calls for further close examination of the processes of democratic reforms, in order to understand when and how institutional constraints structure the debates and the final outcome. Indeed, the case study of Italy and the electoral and constitutional reforms of 2005 have already provided an opportunity to demonstrate that where broad coalitions are involved the ability to come to an agreement on institutional reforms depends on the ability to negotiate a compromise which provides guarantees to all actors in the coalition. [58] This example illustrates Bruno Palier’s idea of “ambiguous consensus”, which postulates that a consensus on reforms is not necessarily the object of a clear agreement on the set objectives. This helps explain why the presence of multiple veto players is not always sufficient to block change. [59] Second, this study has shown that, in the long term, the level of political support is closely connected to the number of reforms adopted. The democracies most affected by the “erosion of political support” diagnosed by Dalton, such as Italy and France, are therefore more vulnerable to democratic reform. Again in the case of Italy, this explains why the question of institutional reform was as pressing in 2014 as in 1993. This article has demonstrated the impact in the short term of an increase in volatility and political alternation on the number of reforms adopted during a given parliamentary term.

52In a context of prolonged crisis, we might therefore expect that circumstances are favourable to an increased number of democratic reforms in established democracies. The level of political support is falling. The Irish elections of 2011 or those in Italy in 2013 recorded unprecedented levels of volatility. In most of the elections held since the crisis in 2009, the parties or coalitions in power suffered heavy electoral defeat. We still need better understanding of the mechanisms that link political support, alternation and volatility through detailed case studies. A further research question directly stemming from these results also springs to mind. If it is the most disgruntled democracies, and those most subject to electoral upheaval, that implement the most reforms, to what extent are these reforms able to shore up political support and limit the impacts of electoral ups and downs?