Determinants of opposition fragmentation

1. Introduction

In September of 2003 in Japan, the Liberal Party, the second largest opposition party, merged into the Democratic Party of Japan, the largest opposition party, which then dramatically increased its seat share in the general election in November. Similarly, in Canada in December of 2003, two opposition parties, the Progressive Conservative Party and the Canadian Alliance, merged to form a new party, the Conservative Party of Canada, which also increased its number of seats in the next general election in 2004 and eventually won power in the 2006 election. In both of these instances, the goal of the merger was to compete with the ruling party – the Liberal Democrats in Japan and the Liberals in Canada – both of which had been continuously in power for a long time. Opposition leaders must have considered that, without a merger, their parties had little chance of defeating the stronger incumbent.

Where two or more opposition parties exist and seek to compete with the ruling party for control of the government, the rational expectation would be a merger that forms either a new opposition party or enlarges an existing one. The cases of Japan and Canada are examples where it happened. Yet, mergers of opposition parties do not happen everywhere, and there are many countries where the opposition is fragmented into many parties. As will be shown later, the degree of opposition fragmentation varies greatly across countries. This study examines the determinants of opposition fragmentation in democracies.

Why is this topic important? It is because whether the opposition of a country is unified or fragmented makes a difference in how the country’s democracy functions. Maeda (2010) demonstrates that the presence of a unified opposition party or electoral alliance facilitates anti-incumbent economic voting, compared with the situation in which multiple opposition parties exist. His reasoning is that a unified opposition is perceived by voters as a credible alternative to the incumbent and thus serves as the point of convergence for those who want to punish the government, making the vote transfer from government parties to opposition parties more straightforward and inducing the desire to punish the government. On the contrary, if the opposition camp is fragmented into many parties, each with a distinctive policy agenda, the choice of the discontented voters who previously voted for the current incumbent will not be clear, and votes against the government necessarily split, rather than converge.

This difference should create important consequences. If the level of opposition fragmentation changes the likelihood of an electoral loss among ruling parties, then the motivation of incumbents to be responsive to citizens’ demands may be influenced by how fragmented the opposition is. A ruling party in a country with a fragmented opposition does not face a high risk of defeat, even during times of poor economic performance. It therefore may not seriously try to deliver what people want and stay popular. If this happens, the accountability of the government will deteriorate. ¹ The degree of opposition fragmentation thus appears to be an important characteristic of democratic governance.

Although there is a large literature on the number of parties and party system fragmentation (e.g. Duverger, 1954; Lijphart, 1994; Sartori, 1976), there has been no systematic research on the determinants of opposition fragmentation. In general, studies on opposition parties have been relatively scarce. In the 1960s, two well-known studies, Dahl (1966) and Ionescu and de Madariaga (1968), were published, and at the time it was stated that ‘very few works have ever dealt with [the subject of opposition] specifically and exclusively’ (Ionescu and de Madariaga, 1968, 2). At the time of their writing, the Fifth-Republic France and West Germany had not experienced an alternation in power; Italy, Sweden and Japan were under one-party dominance; and countries such as Spain and Portugal were not democratic. The state of opposition parties was quite different from what it is now.

The late 1960s and early 1970s were a time when the study on opposition received a reasonable amount of scholarly attention. Yet, the attention subsided after that. Blondel (1997) points out a relative dearth of comparative studies on opposition since Dahl (1973), in comparison to an abundance of case studies. One of the reasons for the decreased attention to opposition would be the fact that, since the 1960s, more and more parties experienced governing, and thus the distinction between government and opposition has become less clear. This trend coincides with the emergence of ‘cartel parties’ (Katz and Mair, 1995; also see Katz and Mair, 2009). According to this view, parties have become ‘semi-state agencies’ and ‘the set of “governing parties” is no longer as limited as it once was’, and thus ‘almost all substantial parties may now be regarded as governing parties’ (Katz and Mair, 1995: 107).

If the cartel party thesis is true, the distinction between governing parties and opposition parties is no longer important, and thus the meaning of studying opposition parties specifically is also small now. However, there have been recent attempts to shed new light on opposition from a theoretical and comparative perspective (see, e.g., Helms, 2008 and De Giorgi, 2007). I argue that studying opposition parties is still relevant. As demonstrated by Maeda (2010) cited above, the degree of opposition fragmentation influences election results, and thus can indirectly affect governments’ performance. Findings from this study will contribute to the debates on the importance of opposition in democratic systems. In addition, as I discuss in the conclusion, this study has valuable implications for the design of political institutions.

2. Parliamentary Rules and Opposition Strategies

Although there can be many factors that influence opposition fragmentation, my primary interest in this study is on the role of parliamentary rules. My principal argument is that parliamentary rules lead to differences in behaviours and strategies of opposition parties and thus opposition party systems. The following causal connection is proposed:

Parliamentary rules → Opposition parties’ motivations → Level of opposition fragmentation

Let us start with the first part of the proposed causal connection. In countries such as the United Kingdom the parliamentary rules allow the party in government to dominate the legislative process, and opposition parties have little, if any, influence on policymaking. In most situations, bills proposed by the government are smoothly passed by the legislature even if opposition parties are opposed to them. In other systems, on the contrary, the parliaments are more deliberative in that opposition parties have influence on policies through the legislative process. These parliaments have various rules that promote deliberation in policymaking – for example, committees with strong powers and committee chairs that are distributed even to opposition parties (see, e.g., Strøm, 1990). Germany is an example of such systems where ‘the government succeeds in getting about 85 percent of its initiated legislation passed by the Bundestag, but it is very often amended and altered in many ways to reflect a variety of specialized interests’ (Powell, 2000: 63). In those systems, parliamentary rules allow opposition parties to influence policies through parliamentary deliberations. Let us use the terms of ‘majoritarian parliament’ and ‘deliberative parliament’ to distinguish the two types described here. In short, deliberative parliaments allow stronger opposition influence than majoritarian parliaments.

I argue that these differences in parliamentary rules shape opposition parties’ motivations and strategies. All other factors being equal, parties prefer to be in power than in opposition. Yet, the extent to which being in power is preferable to being in opposition varies across parties and also depends on the environment in which parties operate. Some parties value their policy positions more than others that are more pragmatic about positions in their quests for power. Environments can modify parties’ orientations because the relative benefits of being in power in comparison to being in opposition are different across contexts. Opposition parties in deliberative parliaments can partially realize their agendas, or at least modify the government’s proposals to their preferred direction, which gives them something to advertise to their supporters as their achievements of being in parliament, albeit not in government. In such situations, supporters of opposition parties may be partially pleased even though their parties are not in power, and they may keep supporting the parties even if the perceived chance of winning power in the future is small. In contrast, opposition parties in majoritarian parliaments have little to gain from their status quo position. Their supporters may be frustrated at their party’s inability to influence policies. Supporters of smaller opposition parties may abandon their parties and switch their support to a larger opposition party that has a better chance to win power and alter the policies implemented by the government. Thus, majoritarian parliaments place opposition parties, especially smaller ones, in a harsher situation than deliberative parliaments do. Opposition parties in deliberative parliaments have something to gain from the status quo while those in majoritarian ones have nothing and are continuously in need to win office. In short, on average, opposition parties’ power-seeking motivations are higher under majoritarian rules than under deliberative rules.

There is also an indirect channel through which deliberative rules lead to higher opposition influence in policymaking. Strøm (1990) has shown that minority cabinets are more likely to be formed in countries with deliberative parliaments with strong committees. Minority cabinets need to negotiate with opposition parties for policymaking, giving a bargaining leverage to opposition, which means a stronger opposition influence.

The argument presented here about the varying motivations of parties’ quest for power may seem odd from the perspective that all parties and politicians prefer more power to less power. Yet, it is not uncommon in the literature to consider that parties have various goals (see, e.g., Müller and Strøm, 1999). Some parties do not necessarily seek to maximize votes and seats and pursue power if doing so necessitates a significant change in identity or platform. Maeda (2012) discusses a dilemma that caught legislators of the Japan Socialist Party in the 1960s between seeking power and maintaining the party’s platform and shows empirical evidence that re-election incentives of the Socialist legislators led many of them to oppose a proposed platform change that would have increased the party’s overall support but have decreased the chance of their own re-election. In general, incumbent legislators want to be re-elected. In many situations, a party’s quest for votes, seats and power helps its legislators’ re-election. Yet, there are also situations in which legislators are better off in terms of their re-election prospects if their party tries to maintain the current size and does not actively seek power. It is thus reasonable to observe varying degrees of power-seeking motivations in political parties.

Let us shift our attention to the latter part of the causal connection proposed at the beginning of this section. Differences in opposition parties’ motivations and strategies create differences in the level of opposition fragmentation. I argue that stronger incentives for power lead to less fragmented opposition party systems.

When does an opposition become less fragmented? It does when larger opposition parties become even larger, smaller opposition parties become smaller or disappear, or multiple opposition parties merge together. An opposition becomes more fragmented when relative sizes of opposition parties become balanced, an existing party splits, or a new party enters into electoral competition. Those changes in opposition party systems can happen for a variety of reasons, but the degree of power-seeking incentives of opposition parties must play an important role. As seen in the examples from Japan and Canada described in the Introduction, mergers of opposition parties are an effective way of competing against the ruling party because the increased size narrows the disparity between the ruling party and its main contender, increasing the perceived chance of an alternation in power. In countries with a majoritarian parliament, the survival of small opposition parties is continuously threatened by the motivation of voters who might view them as incapable of winning power or influencing policies and hence switch their support to one of the major parties. For the same reason, party splits and new party entries are difficult and discouraged under majoritarian rules. All of these lead to lower degrees of opposition fragmentation in majoritarian parliaments.

The same effect can also be explained from a different angle. Norris (2004) distinguishes ‘bonding’ and ‘bridging’ strategies of parties. A party with a bonding strategy tries to establish strong and enduring ties with its supporters. A bridging strategy, on the contrary, is to form a broader and looser coalition across various groups and to gather votes from anyone to whom a party can appeal. I argue that which strategy a party chooses is partially determined by the levels of parties’ power-seeking motivations. On the one hand, parties with a relatively low level of power-seeking motivation are likely to take a bonding strategy since it is a safe and risk-averse behaviour. Incumbent politicians whose primary concern is to maintain their jobs as legislators until their retirement do not want to take a risk but want their party to take a defensive strategy and keep the existing supporters. On the other hand, strongly power-seeking parties will be more willing to take a bridging strategy. Although it is a riskier strategy for a party which can result in alienation of its supporters, a party that gains little from its status quo position will be more willing to accept a risk that is necessary to become competitive against the ruling party. A merger of parties is a risky move that involves a change in parties’ identity, and therefore parties with a bonding strategy are less likely to merge than those with a bridging strategy. If parties in a country have a bonding and defensive strategy, the party system becomes stabilized, which hinders an emergence of a large, unified opposition party.

In summary, it is expected that majoritarian parliamentary rules have effects to decrease the level of opposition fragmentation. Those rules create an incentive structure that facilitates an emergence of a unified opposition party. Even when an opposition becomes fragmented by random events, the rules pressure them to become less fragmented again. Thus, the causal mechanism I am proposing here should be observed both statically and dynamically. The level of opposition fragmentation, viewed statically, is expected to be higher in deliberative parliaments than in majoritarian parliaments. Parliamentary rules are also expected to have a dynamic effect in which majoritarian rules decrease the level of opposition fragmentation in comparison to deliberative rules. The following two hypotheses are tested in the remainder of this paper.

Before proceeding to the presentation of the empirical tests of this hypothesis, the next section explains the method of measuring opposition fragmentation and shows descriptive statistics of that variable.

3. Measurement and descriptive statistics

This study measures the degree of opposition fragmentation by using Maeda’s (2010) ‘effective number of opposition parties (ENOP)’, which applies Laakso and Taagepera’s (1979) mathematical formula for effective number of parties (ENP) ² to opposition parties. So-called ‘support parties’ which do not participate in the cabinet but are in agreement with the government to support it are not included in the calculation. Unlike Maeda (2010), who counts parties that were in opposition when going into elections, this study focuses on parties that were in opposition when post-election cabinets were formed.

The scope of this study is limited to legislative elections that were held between 1945 and 1997 in parliamentary and semi-presidential systems in long-standing advanced democracies. The years of coverage were chosen based on the availability of the data on cabinet composition in Woldendorp et al. (2000). I focus on advanced democracies and exclude democracies in developing countries, most of which are new democracies, largely because the purpose of this study is a theoretical advancement rather than a geographical expansion. There are two additional reasons for this. First, democracies in developing areas, especially new democracies, tend to have unstable party systems, in which political parties are frequently formed and dissolved. Instability in identities of parties poses difficulties in examining changes in party systems. In addition, variables such as the change in ruling parties’ vote share from one election to the next cannot be precisely measured if the party system is too volatile. Second, more practically, the data on parliamentary committee structures in Powell (2000), which are used later in this article, are limited to advanced democracies.

Among advanced democracies, parliamentary and semi-presidential systems are the subjects of this research. The United States and Switzerland are thus dropped. Their political systems are not quite comparable with other advanced democracies, especially in terms of the formation of cabinets, which determines what parties are in opposition. The countries that remain in the sample are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, New Zealand, The Netherlands, Norway, Spain, Sweden and the United Kingdom. Portugal, Iceland and Luxembourg are not included because Powell’s (2000) committee structures variable is not available for them. France since 1958 and Finland are the only semi-presidential systems while all others are parliamentary. Dropping semi-presidential systems from the sample does not alter the results substantively (the results are available upon request).

The dataset for this study contains 277 elections from 18 countries. ³ Figure 1 shows the distribution of ENOP values, based on the numbers of parliamentary seats of opposition parties. ⁴ The figure demonstrates that the degree of opposition fragmentation varies widely. The mean is 2.23 and the standard deviation is 1.16. There are 33 cases where the ENOP value is exactly 1. Those are cases in which only one opposition party existed, and they are mostly found in Australia, Germany and New Zealand. The largest value of ENOP in the dataset is 7.14 for Denmark in 1977. The result of this election is given in Table 1 to illustrate a situation of extreme opposition fragmentation.

OPEN IN VIEWER

Figure 1.

Distribution of effective number of opposition parties (ENOP).

OPEN IN VIEWER

Table 1.

The Danish Parliament after the 1977 election.

Government
	Social Democratic Party	65
Opposition
	Progress Party	26
	Venstre	21
	Conservative People’s Party	15
	Centre Democrats	11
	Socialist People’s Party	7
	Communist Party	7
	Danish Social Liberal Party	6
	Christian People’s Party	6
	Justice Party	6
	Left Socialists	5
Total		175

ENP = 5.17; ENOP = 7.14

Note: The numbers are the parliamentary seats of the parties. The source of the result is Parties and Elections in Europe (http://www.parties-and-elections.eu/), and the source of the Cabinet composition is Woldendorp et al. (2000).

ENOP can also be computed from vote shares of opposition parties rather than seat shares. In the dataset used in this study, the ENOP values based on vote shares and seat shares are highly related (r = 0.95). In the subsequent analyses, I use both measures to test the robustness of the findings.

Since opposition parties are a subset of political parties, the degree of opposition fragmentation should be correlated with the degree of overall party system fragmentation. In a country with two parties, the number of opposition parties should be one; countries with many parties are likely to have more than one opposition party. Figure 2 shows the relationship between ENP and ENOP, both measured by seats. The correlation coefficient between the two variables is 0.61. There are a significant number of cases in which the degrees of overall party system fragmentation and opposition fragmentation are not close, which suggests that the determinants of opposition fragmentation are different from the determinants of overall party system fragmentation and that opposition fragmentation is a variable that warrants investigation in its own right.

OPEN IN VIEWER

Figure 2.

ENOP and ENP.

Some may argue that whether there are many or few opposition parties is simply a function of the size of the cabinet – if many parties participate in a cabinet, few parties remain in opposition. Although the size of the governing parties is indeed a factor that influences opposition fragmentation, its impact is not as strong as some may think. Figure 3 shows the relationship between the share of the parliamentary seats of governing parties and ENOP. It is noticeable that the two variables are negatively related, but not very strongly (r = −0.32). Opposition fragmentation appears to be affected by other factors as well.

OPEN IN VIEWER

Figure 3.

Government’s seat share and ENOP.

Let us also examine the ENOP values by country. Table 2 gives the average ENOP values (based on seats) of the countries listed in descending order. Figure 4 graphs changes in ENOP in each country, and changes in ENP are shown in dashed lines for comparison. Note again that observations were made at elections, and changes in cabinet compositions between elections are not considered. It can be seen that there is a great variation in opposition fragmentation across countries. ENOP values are low and remain low during the observed period in such countries as the United Kingdom, New Zealand (until the 1993 electoral law reform) and Australia. Some countries, on the contrary, had tumultuous histories of changes in ENOP. When a dominant party goes in and out of government, ENOP changes greatly. For example, Sweden’s ENOP dropped abruptly in the late 1970s and went up again in the early 1980s. The drop occurred when the dominant Social Democrats went into opposition after the 1976 election. There was another opposition party, but the Social Democrats controlled 90 percent of the total opposition seats, reducing the ENOP level to 1.22. When the Social Democrats returned to power after the 1982 election, the three parties that ruled the country while the Social Democrats were out of power went back to opposition, increasing ENOP back to the previous level. Similar patterns are also found in Denmark and Norway.

OPEN IN VIEWER

Table 2.

Average ENOP values by country.

Country	# of elections	Average ENOP (seats)
Denmark	21	3.52
Japan	20	3
Italy	14	3
Norway	14	2.92
Sweden	16	2.82
Belgium	17	2.76
France	16	2.71
Netherlands	15	2.57
Finland	14	2.29
Spain	7	2.05
Canada	17	2.05
Ireland	16	1.88
Austria	15	1.42
Greece	8	1.41
Australia	21	1.34
United Kingdom	15	1.25
Germany	13	1.24
New Zealand	18	1.1

OPEN IN VIEWER

Figure 4.

Change of ENOP by country.

The following two sections empirically test the hypotheses of parliamentary rules’ impact on opposition fragmentation. An analysis of the levels of ENOP (hypothesis 1) is first presented, followed by an examination of how ENOP changes (hypothesis 2).

4. Level of opposition fragmentation

4.2. Results

The results of regression analyses are reported in Table 3. Models (1) and (2) are main models, and the dependent variable for (1) is calculated based on the seat shares of opposition parties while that for (2) is based on the vote shares. As the hypothesis predicted, the Committee variable has a positive and statistically significant coefficient, meaning that countries with a deliberative parliament tend to have a more fragmented opposition party system – regardless of whether fragmentation is measured by seat shares or vote shares – than in countries with a majoritarian parliament.

OPEN IN VIEWER

Table 3.

Determinants of ENOP.

	(1)	(2)	(3)	(4)
Dependent	ENOP	ENOP	ENP	ENP
variable	(seats)	(votes)	(seats)	(votes)
Committee	2.648**	2.890*	–0.359	–0.679
	(0.748)	(1.006)	(0.561)	(0.697)
Magnitude	–0.0495	–0.177	0.125**	–0.0201
	(0.102)	(0.136)	(0.042)	(0.080)
Upper	–0.014	–0.028*	0.002	–0.009
	(0.012)	(0.013)	(0.009)	(0.009)
Ethnic	–0.341	–0.652	0.164	0.216
	(0.527)	(0.683)	(0.107)	(0.401)
Magnitude × ethnic	–0.317	0.172	0.245	0.404
	(0.241)	(0.295)	(0.120)	(0.220)
Upper × ethnic	0.006	0.054	0.018	0.047
	(0.068)	(0.067)	(0.049)	(0.048)
Govt fragmentation	–0.227	–0.265	0.090	0.029
	(0.121)	(0.135)	(0.080)	(0.069)
Govt seat share	–0.023**	–0.021*
	(0.007)	(0.008)
ENP before election	0.574**	0.650**	0.782**	0.928**
	(0.088)	(0.105)	(0.081)	(0.088)
Bicameral opposition	–0.362	–0.485	–0.067	–0.170
	(0.268)	(0.318)	(0.138)	(0.192)
Intercept	1.892**	2.314**	0.344	0.759*
	(0.537)	(0.652)	(0.185)	(0.323)
Observations	257	257	257	257
Adjusted R-squared	0.441	0.377	0.810	0.736
SER	0.855	1.034	0.541	0.699

Robust standard errors clustered by country are given in parentheses.

**p < 0.01, *p < 0.05

Models (3) and (4) are shown for comparison. See text for the explanation on why Govt seat share is not included in Models (3) and (4).

The seat share of the parties that participate in the cabinet (Govt Seat Share) has a negative impact, as also shown in Figure 3. The overall party system fragmentation (ENP before election) is positively related to opposition fragmentation, which is expected because ENP and ENOP are related, as shown in Figure 2. Electoral systems and ethnic heterogeneity seem to have a limited, if any, influence on opposition fragmentation.

For comparison, an analysis of overall party system fragmentation or effective number of parties (ENP) – including both ruling and opposition parties – is performed, and the results are presented as Models (3) and (4). The dependent variable is based on seat shares in (3) and vote shares in (4). It should be noted that the Govt Seat Share variable cannot be included in these models. Unlike ENOP, which is based on the result of post-election cabinet formation, ENP is calculated from an election result. In other words, ENP is determined when an election is over, and that is before a cabinet formation process starts. Govt Seat Share is, like ENOP, based on what parties get in the Cabinet, which is determined after an election result is finalized. In short, Govt Seat Share is determined after ENP is determined, and Govt Seat Share is influenced by ENP. It is thus problematic to include Govt Seat Share when ENP is the dependent variable. ¹⁰

It is noticeable that the Committee variable is statistically insignificant in Models (3) and (4) while it is significant in Models (1) and (2). This provides supportive evidence for the previously stated claim that party systems and opposition party systems are influenced by different sets of factors and that opposition party systems deserve investigation in their own right.

5. Change of opposition fragmentation

5.1. Measurement

Let us shift our attention from level of opposition fragmentation to how it changes within a country. As we saw earlier in a Swedish example, ENOP changes when parties move in and out of government; such changes are not considered in this analysis. Rather, my analysis here is concerned with evolution of opposition party systems while the same government continues in office. The stories of opposition mergers in Japan and Canada discussed at the beginning of this article are examples of changes in opposition party systems that took place while the ruling parties did not change. In the case of Japan, the ENOP value based on seat shares was 2.48 before the two opposition parties coalesced; it went down to 1.82 after the merger and further declined to 1.33 after the next election in which the newly merged party performed well while the remaining opposition parties suffered a setback. If the hypothesis of this study is correct, this type of change – a decline in ENOP – should be observed more frequently in countries with majoritarian parliamentary rules than in countries with deliberative parliaments. Hence, the variable of committee structure is expected to have a positive impact on the change in ENOP.

The dependent variable in this analysis is the change in ENOP from one election t to the next t + 1, denoted as ΔENOP. Both seat-based and vote-based values are used. ENOP at t + 1 is computed using the electoral results of the parties that were in opposition between t and t + 1, not the parties that were in opposition after t + 1. ¹¹ As noted above, cases where parties moved in and out of government are not in the scope of this analysis. Operationally, this requires the inclusion of only those observations in which the Cabinet composition did not change and all opposition parties continued in opposition between t and t + 1. ¹² Note that whether or not some of the opposition parties between t and t + 1 became governing parties after t + 1 is not relevant for the selection of observations. The absence of cabinet composition changes between t and t + 1 is the criterion for inclusion, and any changes in cabinet composition before and after the election t + 1 is not an issue here. By this operationalization, ΔENOP measures how the degree of opposition fragmentation changed during an electoral cycle. If a merger of opposition parties took place, if any opposition parties disappeared, or if voters’ support for opposition parties became concentrated toward larger opposition parties, ENOP declines and ΔENOP takes a negative value.

A few hypothetical cases help clarify how I operationalize an opposition party system change. Three illustrative examples are presented in Figure 5. In Case 1, Party A received 60 percent of votes in two consecutive elections, t and t + 1. It was the governing party between t and t + 1 and remained in office after t + 1. In this case, ΔENOP is calculated by measuring the change in the level of fragmentation of the parties that stayed in opposition during the electoral cycle, that is, Parties B and C. At t, both B and C received 20 percent of votes. The ENOP at t is 2. At t + 1, the two opposition parties (B and C) obtained 30 percent and 10 percent of votes, respectively. The ENOP is 1.6. Hence, ΔENOP is 1.6 − 2 = −0.4. A negative value indicates that the opposition became less fragmented during the electoral cycle. Some voters may have abandoned Party C and switched their support to Party B.

OPEN IN VIEWER

Figure 5.

Calculation of ΔENOP.

In Case 2, the ruling Party D decreased its vote share from 70 percent to 40 percent. Opposition party E increased its share from 20 percent to 40 percent, and opposition party F increased from 10 percent to 20 percent. In this scenario, although the vote shares of the opposition parties changed from t to t + 1, the ratio of E to F stayed the same, and the level of fragmentation remained at 1.8. ΔENOP is 0, which means that the level of opposition fragmentation did not change during this electoral cycle. Note that it does not matter whether either – or both – of Parties E and F became governing parties after t + 1 since my dependent variable is calculated by the sizes of the parties that were in opposition between t and t + 1.

In Case 3, a coalition cabinet of Parties G and H was in power during the electoral cycle. There were two opposition parties, I and J, both of which received the same vote shares in t + 1 as they did in t. However, a new party, K, entered the competition in t + 1 and received 10 percent of votes. The emergence of a new party shows that the opposition became more fragmented during the electoral cycle. In this situation, the ENOP at t is 2.27, and it changed to 1.60 at t + 1. ΔENOP is 0.67.

In my sample, ΔENOP based on vote shares has a mean value of 0.009 with a standard deviation of 0.767, and ΔENOP based on seat shares has the mean of –0.009 and the standard deviation of 0.697. The mean values are close to zero due to the fact that it is a variable for the change and takes positive and negative numbers. The standard deviation values show that this variable has a fairly wide variation.

5.2. Analysis

The main independent variable is the legislative committee structure, as was the case in the earlier analysis for levels of ENOP. Control variables that were in the analysis presented in the previous section are included. In addition, the following four variables are specified in the models. First, the ENOP at election t is added (LagENOP). Since the dependent variable is the change in the ENOP values, the base value of the ENOP should be controlled. Second,I control for the change in the vote shares for the governing parties from t to t + 1 (ΔGovtVotes) because whether the opposition parties increased or decreased their total size may make a difference in their fragmentation level. Third, the number of previous elections the governing party (parties) has won is included (Previous Terms). The incentives and strategies of opposition parties may be different depending on how long they have been out of office, and this variable captures the impacts these differences make. Finally, the length of time, measured in months, between the two elections (t and t + 1) is controlled (Electoral Cycle).

The results of the analysis are presented in Table 4. The dependent variable in Model (5) is the change in ENOP based on seat shares, and that in Model (6) is the change in ENOP based on vote shares. The positive and statistically significant coefficients on the Committee variable indicate that opposition parties in deliberative parliaments tend to become more fragmented over time, in comparison to those in majoritarian parliaments, providing clear support for my argument. As expected, the LagENOP variable has negative impacts on the dependent variables. If the opposition is already unified, it is unlikely to become more unified. The variable for the government’s vote change (ΔGovtVotes) also has significant and negative coefficients, which means that if ruling parties’ vote share increases, the opposition becomes less fragmented. This is an intriguing finding that warrants further investigation, but one possible explanation is that parties already small may die or become negligibly small in elections in which ruling parties become even larger. Alternatively, the causality may run in the other direction. That is, when a newly emerging party is increasing its popularity quickly (more fragmented opposition), the government may lose some of its supporters and suffer a defeat (decrease in government’s vote share).

OPEN IN VIEWER

Table 4.

Determinants of ΔENOP.

	(5)	(6)
Dependent	ΔENOP	ΔENOP
variable	(seats)	(votes)
Committee	2.157*	2.348**
	(0.747)	(0.739)
Magnitude	0.016	–0.087
	(0.060)	(0.058)
Upper	–0.006	–0.020
	(0.012)	(0.012)
Ethnic	–0.408	–0.762*
	(0.276)	(0.325)
Magnitude × ethnic	–0.281	–0.177
	(0.183)	(0.129)
Upper × ethnic	–0.019	0.053
	(0.063)	(0.060)
Govt fragmentation	–0.548**	–0.480**
	(0.162)	(0.130)
Govt seat share	0.325	0.853
	(0.610)	(0.663)
ENP before election	0.476**	0.463**
	(0.143)	(0.092)
Bicameral opposition	0.023	–0.102
	(0.076)	(0.112)
LagENOP	–0.498**	–0.400**
	(0.117)	(0.077)
ΔGovVotes	–0.042*	–0.059**
	(0.015)	(0.019)
Previous terms	0.035	0.029
	(0.031)	(0.037)
Electoral cycle	–0.003	–0.002
	(0.003)	(0.004)
Intercept	–0.003	–0.182
	(0.417)	(0.446)
Observations	186	186
Adjusted R-squared	0.213	0.250
SER	0.618	0.664

Robust standard errors clustered by country are given in parentheses.

**p < 0.01, *p < 0.05.

6. Conclusion

This article has sought to identify the determinants of opposition fragmentation and especially how parliamentary rules affect it. The analyses presented above demonstrate that parliamentary rules allowing opposition parties to influence policies are associated with fragmented opposition party systems.

Since the degree of opposition fragmentation affects the electoral fortunes of governing parties (Maeda, 2010), the findings of this article suggest that parliamentary rules may indirectly influence the length of time governments stay in office. Where less deliberative parliamentary rules are used, the opposition party system is likely to evolve into a more unified one, through fusion of parties and/or disappearance of minor parties. A unified opposition poses a stronger threat to incumbents that may shorten their tenure in power. On the contrary, deliberative parliamentary rules (where opposition parties can influence policy decisions without becoming the governing party) tend to foster the status quo among the opposition parties which may make the re-election of the government easier.

If opposition parties are allowed to influence policymaking, it may seem as if the government is doing the opposition a favour. However, the findings here suggest that incumbent governments that allow their opposition some voice in the policymaking process, in addition to appearing more conciliatory and deliberative, may in fact be securing their own survival by giving the opposition more to gain from the status quo. Ruling parties that wish to continue in office must nurture their popularity while at the same time not increase the motivation of the opposition to seriously seek to win power. As I have shown here, opposition parties that are deprived of political influence will reduce their levels of fragmentation in an effort to win power more often than opposition parties that can influence policy and keep their support base relatively satisfied. Thus, governments’ efforts in ‘reaching across the aisle’ may in fact have the outcome of ‘taming the opposition’ by giving them more of a stake in remaining in the opposition.

The above consideration suggests that there may be a trade-off in the design of democratic institutions. Deliberative legislative rules are often considered democratically desirable by scholars and practitioners and are related to the consensus model of democracy which, according to Lijphart (1999: 277), is ‘the “kinder, gentler” form of democracy’. However, the findings of this article imply that legislative rules allowing opposition parties some policy influence can render opposition parties tame and less serious in winning power, which may prolong a ruling party’s tenure. The Westminster model, in which the opposition is deprived of political influence and the legislature quickly passes the executive’s proposals, may seem less democratic to the proponents of consensus democracy, but it may have an advantage in enhancing competition between government and opposition and realizing a regular alternation in ruling parties.