OUR CASE SELECTION

Although Westminster systems have features that make our analysis particularly pertinent—the relative absence of checks and balances on executive power means that the opposition is limited to the use of its voting strategies in order to yield influence—government-versus-opposition patterns are found elsewhere. There are, however, practical reasons of data availability—along with a desire to evaluate the relative, nuanced claims of competing alternatives—to keep the analysis more tightly focused. Although government-versus-opposition voting is a pervasive feature in different countries, it is possible that this empirical regularity evolved independently in each country rather than being related to a common set of institutions. Since government-versus-opposition voting is observed in different places and at different points in time, its emergence may be due to a common set of circumstances that can have an independent effect on the voting patterns in these parliaments. Moreover, even though such voting patterns may plausibly be related to the institutional features of certain democracies, the question remains: which of these institutional features are important determinants of government-versus-opposition voting? To analyze statements about necessary conditions for behavior, we want cases that present a particular pattern of variation on the “right hand side.” Specifically, and for various competing theories, we desire (at least) two cases that differ (only) on the claimed variable of interest, holding other potential confounders constant.

Data from the United Kingdom, the home of the Westminster system, provide just this. The Scottish Parliament shares with the assemblies in Wales and Northern Ireland both timing of its inception and a common, inherited, political culture; thus, we can exploit institutional variation between these countries to explore our hypothesis. To assess our main claim that a majoritarian executive is a requirement for the voting pattern in question, we can compare the Northern Ireland Assembly—which abandons this practice—to that of Scotland. To rule out (or rule in) alternative hypothesis about other institutional factors that may act as driving forces, we can compare and contrast different cases. To assess the effect of legislative domain, we can contrast the Welsh case—which has no budgetary powers—with that of Scotland. If we want to determine whether the behavior in question is driven by electoral considerations, then we can examine the House of Lords, where uniquely, membership is for life. Finally, relative to Westminster, the devolved assemblies are particularly interesting because some MPs are elected directly by their constituents, whereas others are elected from a party list. Thus, we can exploit this within-country variation to assess the importance of different electoral incentives in driving opposition behavior. Performing such analysis provides interesting insights that distinguish the particular institutional basis of government-versus-opposition voting from other factors. As we see, careful case selection allows us to provide concrete evidence that supports our claim that majoritarian control over agenda setting, rather than other institutional factors, explains the pattern of government-versus-opposition voting. We postpone such comparative analysis until the section dealing with alternative theories and the plausibility of our assumptions. In the next section, we introduce our main data for the Scottish case.

THE EMERGENCE OF GOVERNMENT-VERSUS-OPPOSITION IN SCOTLAND

Established in 1998, the Scottish Parliament is the devolved national, unicameral legislature of Scotland, which has 129 elected members each elected for a four-year term. Its ambit includes all Scottish matters not specifically “reserved” to the United Kingdom Parliament; in practice, this means that members may legislate in areas such as education, agriculture, health, local government, and justice. Indeed, ministers in Edinburgh have the same executive powers in these particular policy arenas as their counterparts in London enjoy for their portfolios. A comparison between the behavior of its members with that of members of the Westminster Parliament is instructive. Although most of the members of the newly elected parliament had not previously served in Westminster, the parties they represented were the same as those represented in the House of Commons. This allows us to compare party behavior in both legislatures at the same points in time.

However, there are important institutional differences between the two parliaments that allow us to draw conclusions about the likely cause of any variation in the behavior we find. Unlike its counterpart in London, the Scottish Parliament was designed, both architecturally and in terms of its institutional make-up, to avoid the adversarial nature of politics in Westminster. The crafters of the constitutional reform that brought to an end the period of history since 1707, by which all Scottish affairs were dealt with by Westminster, envisaged a devolved assembly that would work on more consensual lines. For example, the assembly itself was designed so that members of the Scottish Parliament (MSPs) sit in a semicircle rather than on opposing benches.Footnote ⁴ Perhaps more important, MSPs are elected under a mixed member proportional system: 73 constituency members are elected under “first past the post,” whereas a further 56 are elected under proportional representation from party lists in 8 regions. The first two governments comprised a coalition between the Scottish Labour Party and the Scottish Liberal Democrats.

Although the Scottish Parliament deviates from the classic model in important ways, many of the quintessential features of the Westminster democracy are intact. The chamber is unicameral and, although provisions are made for committee and individual members to introduce bills, the lion's share of parliamentary business involves the discussion and debate on measures introduced by the executive (Mitchell Reference Mitchell2000, 610–11). Thus, although the composition of parliament and the rules by which members are elected differ from those found at Westminster, the key feature of Westminster democracy—namely, the fusion of executive-legislative relations with agenda-setting power being given to the majority party—remains central to politics at Holyrood. Similarities in the behavior of the parties across these parliaments can then be related to this common institutional feature. In fact, although avoiding the confrontational politics of Westminster-style arrangements was the aim, a pattern of government-versus-opposition was evident from the outset (see Mitchell Reference Mitchell2000, 620). To show this, we begin by calculating the agreement between party A and party B as the proportion of times the majority of each party vote the same way (“aye” or “no”), dropping all divisions on which one or other of the parties had no members voting.Footnote ⁵ The “disagreement” between any two parties is then calculated as one minus the agreement score.

Figure 1 shows the parties in the three Westminster Parliaments from 1992 to 2005, and the first two Scottish Parliaments from 1999 to 2003 and from 2003 to 2007.Footnote ⁶ Each line represents a parliament, and each point a party in terms of its disagreement score relative to the Labour Party. Because, by definition, the Labour Party must agree with itself 100% of the time, its disagreement is zero, and that party appears to the far left. The party that disagrees with the Labour Party most often in each parliament is assigned to the far right of the line: in this way, the parliaments are easily comparable in terms of relative differences between the parties. The specific numbers listed in the graphic are the actual, absolute, disagreement scores with the Labour Party. Thus, the Conservatives are the furthest from the Labour Party in the 1997–2001 Westminster Parliament, disagreeing 91% of the time. A striking feature of these data is the position of the SNP: broadly aligned with the Labour Party in the 1992–1997 Westminster Parliament, when both are in opposition, but strongly opposed to the Labour Party in Scotland.

FIGURE 1.

Parties Relative to One Another in “Agreement” Terms, Specific Numbers Are Actual, Absolute Disagreement Scores for Parties

Notice relative positions of Tories and Scottish Nationalists at Westminster (first three rows) and Scotland (last two rows), respectively. Con, Conservatives; Green, Green Party; Lab, Labour Party; LD, Liberal Democrats; SNP, Scottish National Party; Soc, Socialist Party.

The evidence suggests that despite their ideological profiles, parties play different roles depending on the parliament in question. Specifically, the Tories appear to be the most recalcitrant members at Westminster, whereas it is the Scottish Nationalists (along with the Scottish Socialists, who do not have seats at Westminster) who are the more antagonistic at Holyrood. These observations fit the notion that during this period the SNP was the main opposition party who opposed measures introduced by the Labour Party-Liberal Democrat coalition. The analysis suggests that the structure of government-opposition voting can be related to the institutional similarities between the Scottish Parliament and the Westminster Parliament.

Of course, although it is well known that “government-versus-opposition” voting is dominant at Westminster (Spirling and McLean Reference Spirling and McLean2007), it is plausible that roll call patterns at Holyrood are dictated by other dynamics. In particular, voting might, in fact, reflect underlying policy divisions. Because we do not observe exactly the same individuals voting in both parliaments and, moreover, because the precise institutional rules differ between parliaments, we need to investigate other factors that may underpin these voting patterns. In addition, our claim that the SNP agrees with the Labour Party in Westminster “x percent of the time” assumes that the parties vote cohesively. It is possible that some members are more or less aligned with their counterparts in other parties.

First, then, we explore the level of intraparty cohesion. In Table 1, we report the “Rice index,” denoted r, which attempts to capture the unity of a party on any particular vote (MacRae Reference MacRae1970; Rice Reference Rice1928). That is, r = 100|p − (1 − p)|, where p is the proportion of the party voting “aye” on the roll call. As can be readily seen, parties at both Westminster and Holyrood average a mean Rice score of more than 95%. Bearing in mind that this includes purported “free votes”, it seems safe to assert that party cohesion is as prevalent and robust in the Scottish Parliament as in the House of Commons. Indeed, with the exception of the Liberal Democrats at Westminster in 1992–97, the median value of the Rice index in every case is 100%.

TABLE 1.

Mean Rice Index Scores for Various Parties at Westminster and Holyrood

^aScottish Socialists consist of one member in the first Scottish Parliament and thus receive a 100% Rice index score by definition. SNP, Scottish National Party.

Establishing that there is one predominant dimension to roll call voting and that dimension is “government-versus-opposition” is slightly less straightforward. We proceed by rank ordering members via the nonparametric optimal classification (OC) routine (Poole Reference Poole2000) and inspect the dimensions in terms of both substantive interpretation and goodness-of-fit. In Table 2, we report the proportion of votes correctly predicted from this procedure and the aggregate proportional reduction in error from the relevant dimension of model.Footnote ⁷ As with all such techniques, these numbers must (almost) always increase as the model becomes more complex; the key is to select a dimension number wherein an additional dimension is not “worth” the extra parameter. In practice, we see that a simple one-dimensional model predicts approximately 95% of the voting decisions correctly for both the first and second Scottish Parliaments.Footnote ⁸ Adding a further dimension allows us an improvement of just 4%. As a comparison, the equivalent figure for the House of Commons (1997–2001) is 99% and for the U.S. Senate (90th), a relatively low 84%. We thus settle for the single-dimensional fit and offer a substantive interpretation of this dimension.

TABLE 2.

Votes in Scottish Parliament Explained in One and More Dimensions

Table 3 reports the rank position (from 1 through 134) of each party by using the party leader's score as a proxy. In terms of Figure 1, the results are similar. Once again, we see the Labour Party and the Liberal Democrats at one end of the continuum, and the Tories and Greens somewhere between them and the Scottish Nationalists and Scottish Socialists. Given that the Labour Party—a traditional party of the trade unions—and the Scottish Socialists occupy opposite poles of the rank order with the right wing Conservatives between them, it is difficult to believe that we are observing a left-to-right phenomenon. An alternative is to regard this dimension as a Scottish independence continuum, with the Scottish Nationalists—who are actively campaigning for full political autonomyFootnote ⁹—at one end and the Labour Party, which opposes such moves, at the other.Footnote ¹⁰ The position of the Conservative Party scuppers this interpretation: unionists by name and nature, they opposed the original creation of the Scottish Parliament. We would thus expect them to appear at one end of the independence spectrum, but here they appear between the Labour Party and the Nationalists on this dimension.

TABLE 3.

Number of seats and Optimal Classification (OC) of First and Second Scottish Parliaments

For each parliament, the first row shows the number of members of (the Scottish) Parliament by party. The numbers in the second row are OC ranks of the leaders of the various parties. Both parliaments have a total of 134 members. MSP, member of the Scottish Parliament.

To summarize, our analysis suggests that parties in Scotland are extremely cohesive in their voting and, in terms of overall patterns of roll call behavior, “government-versus-opposition” is the order of the day. The pattern of government-versus-opposition voting is very different than what we would expect if MPs were voting along ideological lines, and this suggests that this voting is strategic. Moreover, the fact that there is institutional variation between the Westminster and Holyrood parliaments implies that this strategic aspect is related to the common institutional features of these parliaments. The key feature that links them is that the government agenda is implemented by the majority party, which operates without checks and balances.

A MODEL OF WESTMINSTER DEMOCRACY

The data suggest that a pattern of cohesive government-opposition voting is unlikely to arise due to preference-based voting alone and that a significant aspect of government-versus-opposition voting relates to strategic considerations that arise given the institutional infrastructure of Westminster democracies. The key institutional feature related to such voting patterns is the near monopoly of agenda-setting power enjoyed by the executive.Footnote ¹¹ To provide microfoundations, we craft a simple “bare bones” model that captures this core element of Westminster democracies, combined with strong party discipline, and show how these ingredients relate to the voting strategies deployed by government and opposition forces.

Parliamentary Agenda and Party Structure. We begin with a parliamentary agenda that consists of a set of J independent policies. We assume that the policy space for each j ∈ J is unidimensional. As already stated, the main institutional feature of Westminster democracy is that the government agenda is implemented by the majority party. For most of our results, we concentrate on the case where the agenda setter on policy j is the median of the governing party on that policy, but the exact identity of the government representative who holds the agenda-setting power is of no consequence to our results. We write the proposal of the agenda setter on the jth policy as ψ_j.Footnote ¹²

We make a simple adaptation to the agenda-setting model of Romer and Rosenthal (Reference Romer and Rosenthal1978), which has been used to analyze majority party control over the agenda in congressional systems under different procedural rules (Cox and McCubbins Reference Cox and McCubbins2005). Analysis of these models focuses on the unique subgame perfect Nash equilibrium in which, given a proposal made by the agenda-setter, all players vote sincerely. The game we analyze is, in contrast, more suited to the Westminster scenario, where the government majority has strong agenda-setting powers; can block opposition amendments and curtail parliamentary debate via guillotine measures and programming of government bills; and, critically, where parties can commit to their voting strategies. As noted at the beginning of the article, empirical scholars of parliamentary settings have viewed the ability of parties to discipline their members as a distinguishing feature of such systems. Cox's authoritative study, for example, shows that party discipline emerged in the United Kingdom in the 19th century in tandem with increasing control over the agenda exerted by the governing majority. Our framework allows us to explore the implications of this commitment ability alongside the agenda-setting powers of the government. The critical aspect of our timing is that we assume that MPs can choose whether to bind themselves (e.g., invoke the party whip) to voting together before the proposal is made. This is, of course, equivalent to opposition members casting their votes before the government introduces its bill to the floor. In short, we treat the party whip as a “bridge-burning” device that is central to policy outcomes in Westminster systems.Footnote ¹³

To capture the basic party division in Westminster democracy, we begin by assuming a set m of MPs who belong to a governing party and a set n who belong to an opposition party. We assume that the governing party has a majority (i.e., m > n) and that parties are jointly exhaustive and mutually exclusive: all members belong either to the governing or to the opposition party, and no member belongs to both. We define a member as belonging to a party when, on average, her views on a particular policy coincide with those of an (average) member of her party. Thus, parties are formed of like-minded individuals. More formally, we assume that on each policy issue, the ideal points of members of the same party are independent and identically distributed (iid) draws from a (possibly common and known) continuous distribution. For example, for some of our simulations later in this article, we treat the ideal point of each member of the governing party on the jth policy as a draw from a normal distribution with the same mean and variance, and a member of the opposition party has a draw from a normal distribution with a different mean and (possibly) different variance. This assumption means that whereas ex ante backbenchers are homogenous, ex post they are heterogenous with respect to their ideal points on a particular policy. The uncertainty in payoffs is itself not important—what is important is that members can bind themselves to voting with their party once preferences on the real line are revealed (but before the proposal is brought forward).Footnote ¹⁴ However, the variance in members' ideal points across policies captures an important element of Westminster democracy, namely, that on some policies members of the opposition may find themselves more aligned with the government than with their fellow partisans. Somewhat surprisingly, and as we show, this feature of Westminster life is a critical ingredient toward cohesive government-opposition voting.

Voting Strategies. Given the parliamentary agenda, a voting strategy is a vector of votes for the ith member where the jth element of the voting vector corresponds to a binary decision—either “aye” or “no”—on a division over policy j. Some voting strategies are of immediate interest. With “sincere” voting, member i, with “ideal point” x_ij on the real line that forms the policy space of policy j, compares her utility from the jth status quo, which we write as x^q_j, with the utility she receives from the jth proposed bill. Assuming that utility decreases monotonically in the distance between a member's ideal point and the alternatives, then the member votes for the proposal if and only if , where the weak inequality implies the imposition, with no loss of generality, of a tie-break rule whereby indifferent legislators back the proposal over the status quo. Another interesting case is where a member of the opposition votes “no” irrespective of whether this critical inequality is satisfied. This suggests that a member of the opposition strategically opposes a proposal that, in the event that it receives majority support, would provide her with a payoff higher than that attained when the status quo remains in place. The question we ask is: under which circumstances will MPs bind themselves to voting together? In the case of the opposition, such binding commitments can take a very stark form: they commit to voting against any proposal the government brings to the floor. We refer to such voting as “opposition” voting. Our main focus is on situations where MPs can bind themselves in the event where all members from their party are better off by so doing. That is, they bind themselves to a vote on a particular bill when the payoff to all members from doing so exceeds that attained from following a different protocol. Since all must agree to invoke the commitment device, a natural focus is on the Pareto-dominant Nash equilibria of the subgame that involves the agenda setter bringing a proposal to the floor before MPs cast their (final) votes.

Asymmetric Commitment versus Government Agenda-setting Power. Our framework allows us to explore the implications of party commitment alongside the agenda-setting powers of the governing party. We begin by establishing our main results for the case where the opposition has greater ability to commit to its voting strategies. In this sense, we think of commitment as a substitute mechanism that the opposition has available when the government is granted agenda-setting powers. As motivation for this example, we might consider that the government, which maintains a majority, must appeal to a broader coalition of voters than the opposition, which, by definition, is smaller in size and thus more compact. This would be the case, for example, where, as in the UK, many districts are partisan and won by landslides, whereas others are more closely fought contests. When the majority party has a relatively large share of MPs elected in close-run elections, they have less ability to discipline their members. The opposition, in contrast, when reduced to its core, has greater control over the actions of its representatives. We capture this distinction in stark form by assuming that the opposition can commit to a voting strategy, whereas the government backbenchers only ever vote sincerely. This serves as a useful theoretical exercise that helps us establish an interesting baseline result that shows that the ability of the opposition to commit to opposing government proposals can counteract the agenda-setting power of the government.

We illustrate this effect with a simple example in Figure 2 that depicts a single policy agenda: j = 1, m = 3, and n = 2. For this example, we order and label the members of the government by their ideal points g ₁ < g ₂ < g ₃, and do likewise for the opposition so that o ₁ < o ₂. We assume that g ₂ is the government agenda setter. The dashed line indicates ψ_j, her optimal proposal conditioned on different voting strategies deployed by the opposition, and the status quo SQ. The top row illustrates the optimal policy when the opposition opposes all government proposals, whereas the bottom row indicates the optimal policy when the opposition vote sincerely. Under an opposition voting strategy, the proposal must make the (right) pivot with ideal point g ₃ indifferent between her voting strategies. In contrast, under sincere voting, the proposal must leave the median member o ₁ indifferent between the status quo and the proposed policy. It is clear by inspection of Figure 2 that “opposition voting” leaves both opposition members better off than sincere voting. Moreover, given this opposition voting strategy, the government members can do no better than vote sincerely. Thus, although for these parameter values both strategies would be Nash equilibria of the subgame in which the government makes a proposal and MPs cast their votes, here the opposition would choose to bind themselves to voting against any government proposal.

FIGURE 2.

Five-member Example

In both scenarios (rows), g2 is the (government) agenda setter. The first row shows the policy outcome (dashed line, ψj) under opposition voting. The second row shows the outcome (ψj) under sincere voting. Opposition members, o1, o2, are better off in that the final bill passed is closer to their ideal points, when strategically opposing.

There are two critical features of this example. First, the status quo policy lies on the “opposition” side of the agenda setter's ideal point. In Figure 2, the agenda setter belongs to a “left” governing party, and the status quo is to the right of her ideal point. Clearly, when the status quo is to the left of the agenda setter's ideal point, then she can propose her own preferred policy and thus receive the support of her own backbenchers irrespective of the actions taken by the opposition. The second feature is that there is overlap in the ideal points of the government and opposition members. In Figure 2, the ideal point of the rightmost member of the governing party is to the right of the leftmost member of the opposition. When this is not so, then again the votes of the opposition are irrelevant to the final outcome because the agenda setter need only appeal to the government backbenchers. In Lemma 1 in Appendix A, we show that this result generalizes to all such policies.

Here, we focus on the main intuition. On any given bill, with a given status quo and government proposal, an opposition member can find herself in one of two situations: either (1) her vote is irrelevant to the final outcome because, given the government proposal, a majority of government members vote “aye”; or (2) her vote can influence the final outcome, and as a member of the opposition bloc, she votes against the proposal. The latter situation may arise when she finds herself in agreement with the government proposal under sincere voting, whereas some members who belong to the government majority are at odds with the proposed policy. In this situation, the optimal policy proposal (under sincere voting) makes the member strictly indifferent between voting for or against the government. If, in contrast, she strategically votes “no,” then the government must satisfy its recalcitrant government members. Such strategic opposition can induce a proposal that is more favorable to all opposition members; thus, if they could (and they can), then opposition members would bind themselves to voting in this way.

This example provides an illustration of behavior consistent with the stylized fact of cohesive government-versus-opposition behavior: the opposition votes cohesively; the government is cohesive; and, although the government supports the agenda setter's motion, the opposition does not. Several other features are interesting: the opposition votes against proposals that leave them better off than the status quo; and, finally, there are policy consequences of the opposition's action—the outcome is closer to the median preference of parliament than it would be under sincere voting. The more general insight is that strategic opposition can induce a proposal that is more favorable to all opposition members. In short, opposition voting counteracts the government's agenda power.

In our discussion, we have focussed on the comparison between “opposition” and “sincere” voting on a particular bill. As we have shown there are bills on which, if the opposition were faced with a choice between these options, they will bind themselves to use of the former. Of course, we must also consider other voting strategies—it is possible that there is another option between outright opposition and sincerity that yields a better outcome for all—and we do so in Proposition 1 in Appendix A, where we state our result more formally. There, we show that the equilibrium of our voting game for the entire parliamentary agenda, with asymmetric commitment, involves the opposition adopting a strategy that is a convex combination of “opposition voting” on some bills and “sincere” voting on others; in the former, the opposition binds its members to voting against any proposal; in the latter they are free from party discipline. We also show (Proposition 2) that for some range of the parameter space, this equilibrium is robust to deviations by coalitions of opposition members (technically, it is coalition proof).

Symmetric Commitment. Thus far, we have considered a situation where the opposition has greater commitment ability than the government. Although the government can bring take-it-or-leave-it policies to the floor, the opposition can then combine to counteract the government's agenda-setting power. As seen, the party whip is a particularly powerful tool for the opposition to wield because it brings them policy concessions that, in the absence of such a commitment device, they would be unable to attain. In light of this, it is useful to consider a world where the government has access to a similar commitment technology that it can use to buttress its agenda-setting power. We suppose that the government can also commit to its voting strategies in the event where, doing so, they make all members better off. The equilibrium of our game is robust to this adaptation. In short, regardless of whether the opposition chooses to bind itself to opposing all government policies, all combinations of proposals and government voting strategies must make at least one member of the government worse off than they would be under under sincere voting. As a result, the government cannot bind itself to do other than vote sincerely. We state this result formally in Proposition 3 in Appendix A. The intuition is, however, immediate from inspection of our example depicted in Figure 2. There, recall that the opposition opposes all government proposals. Then the final outcome lies within the Pareto set of government MPs ideal points. Thus, although commitment yields a Pareto improvement for opposition members, it can yield no such improvement for the government. Cohesive opposition voting underpinned by the development of stringent whipping can then be viewed as a prudent strategy in response to a government with agenda control that cannot best respond by using similar strategies. A conjecture is that such opposition strategies provide incentives for the governing executive to increase centralization of the legislative initiative (see Cox Reference Cox1987, ch. 6).

Other Forms of Commitment. We have yet to consider what happens when the government has commitment ability and the opposition has none, or when the government can commit to other voting strategies that do not yield Pareto improvements over those obtained by sincere voting. As seen, the strength of opposition voting hinges on the fact that, when the opposition votes against all government proposals, the government must appeal to their recalcitrant backbenchers by making a policy concession to them. Suppose, however, that the government could obtain a “yes” vote from some such members on policy proposals that, if realized, leave these MPs worse off on that particular bill than they would be under sincere voting. For example, returning to Figure 2, suppose that the agenda setter could obtain a policy closer to g ₂. As shown by Huber (Reference Huber1996) and Diermeier and Feddersen (Reference Diermeier and Feddersen1998), if the government can credibly invoke the threat of a confidence vote procedure, under which they might lose their agenda-setting power, then they can extract policy concessions from their backbenchers.Footnote ¹⁵ If, on a given bill, the government has credible access to such a procedure, then, in our setting, the opposition would no longer be able to extract any gains through the use of their voting strategies. They may well oppose (or perhaps not), but either way there are no policy concessions to be had from so doing.

Opposition Voting in a Large Parliament. Our results show that a strategy involving outright opposition on some bills can yield dividends to the opposition. Here, we exploit a key element of our model—namely, that, on average, members of the opposition are in agreement with each other; thus, the average gains from use of their voting strategies should be similar. We explore how the magnitude of this advantage varies with different exogenous parameters: the distribution of ideal points, the size of the assembly, and the relative size of the government.Footnote ¹⁶ We proceed by making distributional assumptions on these key parameters and simulating roll call behavior in large parliaments.

Overlap and Strategic Advantage. Our main results confirm that the relationship between polarization and the advantage of strategic voting is nonmonotonic. In particular, if some members of the opposition are closer to the government agenda setter than to members of the government who would be decisive if the opposition vote strategically, then the incentives for strategic voting are undermined. Why is strategic voting dominant only when polarization takes intermediate values? The intuition is straightforward. If polarization is high, then the opposition is largely irrelevant to outcomes, and members of the opposition are indifferent between their voting strategies. If it is too low, then at least some opposition members find themselves in relative agreement with the government agenda setter and prefer to vote sincerely. We illustrate this result in Figure 3, which simulates a large parliament where 301 members belong to the governing party and a range belong to the opposition. On each policy j, the ideal points of the governing party members are drawn from a normal distribution with mean −1 and variance 2, whereas all opposition members’ ideal points are drawn from a normal distribution with a mean that varies between 7 and 11 at regular intervals. A status quo is generated as a single draw from a uniform distribution with a given maximum and minimum on the real line. For each mean value in the sequence, a total of twenty parliaments are simulated, each consisting of 1,000 bills.

FIGURE 3.

Simulation Results, Assuming Government of Fixed Size (301 Members)

Curves show utility gain of opposition/strategic voting (over sincere voting) for most moderate member (most “left wing”) of opposition party, for various numbers of opposition MPs and overlaps. The line labeled “300 MPs” refers to an opposition of size 300. Below 0 on the y axis, there are negative consequences of strategic voting: in these scenarios, the most moderate opposition member is better off using sincere voting. MP, member of parliament.

On the horizontal axis, we plot a measure of polarization that we term “overlap.” This is the proportion of opposition members whose ideal points are to the left of the rightmost member of the government divided by the size of the parliament. When this measure equals zero, the set of ideal points of government and opposition members are disjoint. In contrast, if overlap is 1, then, on average, there is no ideological distinction between the views of members of government and opposition. The (smoothed) curves on the vertical axis depict the payoff differential to the most moderate (most “left wing”) member of the opposition between adopting a sincere voting strategy and an opposition one when the proposal is endogenous to these voting strategies.Footnote ¹⁷ When the curve lies above zero, on average, the payoff of an opposition strategy for the most moderate member of the nongoverning party, and therefore all other members of the opposition, is higher than that of a sincere voting one. Each curve relates to a different size of opposition ranging from 100 to 300. The shape of the curves depicting the payoff differential follows the logic described previously: starting at zero overlap (complete polarization), the payoff differential is increasing, and it reaches a maximum at intermediate levels of polarization.

Strategic Voting and Policy Moderation. A related question concerns the impact of different voting strategies on the policy outcomes of the parliament. Under sincere voting, upon conclusion of the parliamentary agenda, and on each policy, the outcome will lie in the set of points between the ideal point of the agenda setter and the median member. Under strategic voting, all outcomes lie in the set of points bounded by the ideal point of the agenda setter and the government pivot. In both situations, the agenda setter exploits her role to induce outcomes that are as close as possible to her ideal point, and so the Condorcet outcome is chosen only in the rare case where the agenda setter has the median preference. A corollary of our main result is that if the opposition vote strategically, then outcomes can be closer to the outcome that would prevail under sincere voting. To illustrate, we simulated a large parliament as previously shown. We recorded, for every bill, the distance between the ideal point of the median legislator and the outcome under opposition voting minus the distance that emerges under sincere voting. We then took means. When this (mean) difference is negative, opposition voting yields outcomes closer to the median legislator than sincere voting does. Conversely, when this difference is positive, the outcome from opposition voting must be further from the median legislator (and thus the distance further) than that under sincere voting. Figure 4 displays our results for different opposition sizes and demonstrates a nonmonotonic relationship: the welfare improvement due to “opposition” voting initially increases with polarization, before decreasing as “sincere voting” becomes beneficial.

FIGURE 4.

Simulation Results, Assuming Government of Fixed Size (301 Members)

Curves show moderation effect for various numbers of opposition MPs and overlaps. Below 0 on the y axis, curves imply that “opposition” voting has a moderating effect; above zero, curves imply that “sincere” voting has a moderating effect. MP, member of parliament.

Related Notions of Opposition. We end this section by briefly discussing how our results relate to other approaches to cohesion. An observational implication of our findings is that individually rational MPs cast in the role of opposition should vote as a bloc. In our account, they do so because, by cohesively opposing the government, they counteract its agenda-setting power. Our noncooperative and policy-based approach provides an alternative to that found in the cooperative literature. Straffin (Reference Straffin1997) adapts the Shapley index to show that a minority can increase their voting power through bloc voting. However, if the majority best responds (by also voting as a bloc), then the minority would exert more power (in the Shapley-Shubik sense) if there was no bloc voting. Of course, there are other theoretical reasons for bloc voting: Cox (Reference Cox1987, 66), argues that whipped votes were popular precisely because they provided MPs with a strong cue.

Discussion. Although cohesion has previously been understood as arising due to internal party pressures, we have shown that the distinct pattern of government-versus-opposition voting is due to strategic interaction between the backbenches. Opposition parties can use their votes to counteract the agenda-setting power enjoyed by the government. In particular, when the government has the power to make take-it-or-leave-it proposals to the floor of of the House, the opposition can extract policy concessions by adopting a stance of outright rejection of any government proposal. Such opposition, in turn, reinforces government cohesion. A surprising corollary to this result is that policy outcomes can be closer to the median preference of the parliament under strategic rather than sincere voting. The ability of the agenda setter to move outcomes away from those that would be implemented under the Condorcet procedure (in a unidimensional policy space) is mitigated by opposition strategies. Our analysis suggests that Westminster-type systems may not be as unrepresentative, or the role of opposition party members as marginal, as previously believed.

ALTERNATIVE THEORIES AND PLAUSIBILITY OF OUR ASSUMPTIONS

Thus far, we have shown that the basic pattern of government-opposition voting is consistent with what we would expect to find when the opposition adopts optimal voting strategies in Westminster democracies. New data from the Scottish Parliament and our theoretical analysis point to the conclusion that the pattern of government-opposition voting is related to the basic institutional features of Westminster democracies and, in particular, to the agenda-setting power of the majority party that forms the executive. In this section, we check the robustness of our findings by analyzing our underlying assumptions and by taking on board alternative institutional accounts of our data. In particular, we address the possibility that government-versus-opposition voting arises due to (1) electoral signaling, (2) incentives due to the electoral system, and (3) the different jurisdictions of the parliaments in question.

Analyzing Our Assumptions. Our results rely on two main assumptions: the ability of the opposition to credibly commit to their voting strategies, which we have already discussed; and the monopoly over agenda-setting power enjoyed by the government. Our key results depend on a further condition: namely, the existence of overlap in the ideal points of MPs.

Agenda-setting Control. Our analysis suggests that the opposition will adapt its voting strategies in the face of a government with powers to bring take-it-or-leave-it offers to the floor. Doring (Reference Döring2001) uses measures related to three indicators to assess the agenda-setting power of European national governments: the extent of the government's prerogative over bills, the relative autonomy of committees during parliamentary passage of a bill, and the structure of the parliamentary phase of budgeting. Among three countries deemed to have high agenda-setting powers is the United Kingdom, the paradigm case of the Westminster system (the others are France and Greece).Footnote ¹⁸ In Westminster systems, we find relatively few checks on executive power, and so agenda setting is a particularly potent weapon for the government. However, even in the absence of regional or other power centers from where they can wield influence, the opposition parties can still extract concessions via the voting strategies they adopt in the lower chamber. In fact, our analysis offers a different interpretation of one of the main findings in Doring (Reference Döring2001), namely, that countries that “rank highest in government control of the agenda, obviously pass more conflictual laws,” where the conflictual nature of a law is determined by the proportion of “no” votes. Our analysis suggests, instead, that opposition voting is a direct strategic response to the government's agenda-setting power, is not conditioned on the type of bills the government proposes, and can lead to greater policy moderation.

The Overlap Condition. A necessary condition for strategic opposition voting is the existence of overlap in the ideal points of government MPs and those of opposition members. This critical overlap condition is a very weak requirement. Nevertheless, empirically identifying this effect is not straightforward. This is because analysis of roll call votes conflates the ideal preferences of MPs with the strategic incentives they face to vote in particular ways. In particular, such votes reveal the pressures for conformity induced by the party whip. That said, we can look at a series of ostensibly “unwhipped” votes—whose status was reported as such in official Commons documentation (see House of Commons Library 2008)—and estimate a one-dimensional spatial model from them for all parties. We give details in Appendix D, but our key finding is that for 1997–2001 and then 2001–2005, approximately one third of the opposition party's MPs are located to the left of the most right-leaning government MP. Hence, to the extent that we can treat these roll calls as valid indicators of sincere latent ideological positions, overlapping preferences are an empirical fact.

Alternative Theories. We consider three alternative accounts of the government-versus-opposition pattern in our data.

Electoral Signaling. We begin by dealing with a (we believe) common view that cohesive government-opposition allows voters to demarcate parties in policy terms. That is, parties adopt such strategies to signal to the electorate that there is, in fact, “clear blue water” between government and opposition on the issues of the day. That is, although the opposition might prefer a particular government to the status quo, it still votes against it to signal general opposition and “difference” to the governing party.Footnote ¹⁹ Although the clear blue water hypothesis is, we believe, commonly accepted, it does not provide a convincing account of our data. In elections to Westminster that took place in Scottish constituencies, the main two-party races—in both 1992 and 1997—were between the SNP and the Labour Party. The same is true of elections to the Scottish Parliament. Thus, the clear blue water argument cannot plausibly explain both the Westminster and Holyrood agreement scores observed in Figure 1.

A more compelling argument against the electoral signaling account comes from analyzing data in assemblies that share the basic institutional features of our model but where electoral incentives are absent. Members of the House of Lords are not elected and are appointed for life. As a result, they have no individualized electoral incentives to vote strategically because they have nothing to gain from establishing a difference in policy position between their party and another. Figure 5 displays the results of a one-dimensional OC of the roll call data for the House of Lords in 2001–2005. As seen, the Labour Party is generally to the “left” of the cross-benchers and Liberal Democrats, who are to the “left” of the Conservatives. However, the identities of the ranked peers are instructive: Baroness Amos and Lord Williams—both leaders of the House of Lords during the period—appear to the “left” of their party (actually, they share the same rank). Meanwhile, Labour Party “rebels” such as Baroness Kennedy of the Shaws (see Russell and Sciara Reference Russell and Sciara2007, 307), and Lord Plant appear to the right of these government office holders. Of course, this makes little sense from an ideological perspective, yet is entirely in keeping with what we observe in the House of Commons (Spirling and McLean Reference Spirling and McLean2007): the one-dimensional classification is simply a rank order from most “progovernment” to most “antigovernment” (note that Margaret Thatcher, former Tory prime minister and Lord Strathclyde, the head of the opposition, appear far to the “right” on this plot).Footnote ²⁰ The bottom line here is simple: with unified agenda control, but in the absence of individualized electoral incentives, voting takes a government-versus-opposition form.Footnote ²¹

FIGURE 5.

OC for House of Lords

Note that government office holders appear to the “left” of the Labour Party, whereas rebels appear to the “right” of the same party. This is as expected for government-versus-opposition voting. OC, optimal classification.

The historical record is also more consistent with our account. As early as 1766, contemporary parliamentarians such as Lord Rockingham were able to classify their colleagues in the House into a “government” and “opposition,” along with an approximately equal number of independents (Namier and Brooke Reference Namier and Brooke1964, 190–91). The opposition during the period was not a unified force, but rather a series of personalized (Whig and Tory) factions lead by aristocratic notables. Note that this is some 60 years before the first Reform Act of 1832, and thus there can be no notion that the parties—such as they were—offered alternative programs to woo voters in the constituencies. Indeed, relatively few districts were competitive in any sense: “rotten” boroughs—wherein the electorate was utterly controlled by local landowners—were common and typically elected two members to the legislature. Even as late as 1820, more than one third of the 558 MPs owed their positions to patronage rather than open contest (Thorne Reference Thorne1986, 44). Thus, a pattern of roll call behavior consistent with a government-versus-opposition divide substantially predates competitive elections and parties with well-defined and ideologically distinct appeals to voters.

Incentives Due to the Electoral System. As noted in our previous discussion of the Scottish case, the Holyrood Parliament differs from Westminster in that some Scottish Members of Parliament (SMPs) are elected under list Proportional Representation (PR). The fact that the list system appears to increase parliamentary representation for certain groups (Brown Reference Brown2000) might provide different incentives and lead one to expect differences in the voting behavior of MPs elected under this system. It is then plausible that differences in the behavior of elected Scottish representatives in Whitehall and Holyrood can be explained by differences in the electoral system. As a fillip to our argument that it is the institutional features common to both Westminster and Holyrood—namely, the control over the agenda that is enjoyed by a majoritarian executive—we compare the voting records of those SMPs elected under single member district (SMD) plurality, 73 in all, to that of the 56 members elected under list PR. In fact, and as explained in Tables 4 and 5 in Appendix B, we fail to reject the null hypothesis that the ideal points of the members from these different samples are drawn from the the same distribution; thus, there is no obvious difference between the behavior of list and SMD legislators within the same party.

TABLE 4.

Kolmogorov-Smirnov Test for Scotland

Null hypothesis is that list and SMD assembly members' estimated ideal points are drawn from the same distribution. A p > .05 implies that we fail to reject the null (i.e., list and SMD members behave similarly). SMD, single member district; SNP, Scottish National Party.

TABLE 5.

Kolmogorov-Smirnov Test for Wales

Null hypothesis is that list and SMD assembly members' estimated ideal points are drawn from the same distribution. A p > .05 implies that we fail to reject the null (i.e., list and SMD members behave similarly). PC, Plaid Cymru; SMD, single member district.

Differing Jurisdictions. A plausible hypothesis is that patterns in our data reflect the functional separation of policy domains. For example, although Holyrood and Westminster are similar in the structure of the executive, the jurisdiction of the parliaments is markedly different. To explore whether policy domain or executive structure can best explain the pattern of government-versus-opposition voting, we look beyond these parliaments to the assemblies of Wales and Northern Ireland. These assemblies are very similar in terms of their jurisdiction. They also share, to a large extent, a common inherited political culture that may have an independent effect of patterns of legislative behavior. However, and as we describe later, these assemblies are very different with regard to the structure of the executive.

Initially, the powers of the Welsh Assembly were simply those previously held by the Secretary of State for Wales, and there was no separate executive. In 2006, the new Government of Wales Act formally divided the legislative and executive, with the latter drawn from the former. But, even with the earlier arrangement, it was the Cabinet—rather than the committees—that played the key role in developing and introducing legislation for consideration (Laffin and Thomas Reference Laffin and Thomas2000).Footnote ²² Hence, the agenda-setting role we allude to in our model holds for the executive in Wales.Footnote ²³

In contrast, the Northern Ireland Assembly, although similar in terms of policy domain to the Welsh Assembly, was designed with “consociational” (in the sense of Lijphart Reference Lijphart1977) principles in mind, and the legislature has various features that make it very different to the House of Commons. In Northern Ireland, in contrast to Westminster, Scotland, and Wales, agenda-setting power is not controlled by a majority. The Cabinet is necessarily drawn from both unionist and nationalist communities with specialized procedures for allocation of portfolios (see O'Leary, Grofman, and Elklit Reference O'Leary, Grofman and Elklit2004). Following the 2007 election, the parties included in government were represented by more than 90% of the members. Moreover, in Northern Ireland, the planned roll call agenda is subject to special voting procedures that require “parallel consent” and “weighted voting” to account for preferences on both sides of the constitutional divide (Wilford Reference Wilford2000, 581). This is combined with a strong (and diverse by composition) committee system and a mutual veto for some matters. In short, in Northern Ireland, in the absence of majority control of the agenda, and with several checks and balances on executive power, we would not expect strategic “opposition” voting at division time.

Figure 6 presents the OC of Welsh roll call votes for the first (1999–2003) and second (2003–2007) meetings of the assembly.Footnote ²⁴ A single dimension correctly predicts more than 97% of the votes in both periods. On this dimension, Plaid Cymru, the Welsh nationalists, appear to the “right” of the ruling Labour government. This cannot represent an ideological scaling because Plaid Cymru are to the left of the Labour Party and the Liberal Democrats on socioeconomic issues, and furthest from the Conservatives on matters of constitutional reform. In fact, their placement is as we would expect according to government-versus-opposition voting, and this is in keeping with qualitative accounts of recent Welsh parliamentary history (McAllister Reference McAllister2000). A further justification for our claim is given by examining the specific position of John Marek, a left wing Labour rebel who was deselected by his party after criticizing its policies as being unduly conservative. Notice that Marek moves “right” in the OC plots: a relocation that makes sense only if we regard the scaling as reflective of government-versus-opposition. That is, Marek's true ideological position is to the left of the Labour Party; yet, his scaled position is to the right because he now opposes its proposals. The Welsh case thus confirms that strategic opposition voting emerges in legislatures outside Westminster and Holyrood, given the basic institutional apparatus of an executive able to control the agenda (and some mild conditions imposed on preferences of members).

FIGURE 6.

Scaling of the Welsh Assemblies: Government-versus-opposition

Note position of John Marek: his left wing views cause him to leave the Labour party; yet, he is subsequently positioned to their right. Parties are Labour, Plaid Cymru (Pd Cymru), Liberal Democrats (Lib Dems), Conservatives (Cons'tive), and Others.

Figure 7 shows the data from the Northern Ireland Assembly. The first subfigure refers to an OC of roll calls during the period 1999–2002, after which the assembly was suspended. The second subfigure is then the period from the 2007 elections through to June 2008. In the prior period, the Ulster Unionists (UUP) and the nationalist Social Democratic and Labour Party (SDLP) were the largest two parties and earned the first and deputy first minister slots. In the next period, the Democratic Unionist Party (DUP) and Sinn Féin (SF) took those roles. The other major parties were also, by constitutional construction, included in the Cabinet. If our analysis is correct, then we should no longer expect (larger) parties to consider themselves out of power at any particular time, and in the absence of “opposition,” we would expect parties to vote along ideological lines. Figure 7 demonstrates this pattern: in the upper portion, the coalition of the UUP and SDLP appears in the middle of the one-dimensional scaling (which correctly predicts 95% of the votes), whereas the DUP and SF (who did not possess the top positions at this time) disagree more often with the coalition and, crucially, one another. In the second plot, we see the same pattern with the requisite identities reversed: now the DUP and SF form the center of the dimension (again, a single dimension predicts 95% of the votes correctly). Thus, although the pattern we observe in these figures reflects subtle differences in the make-up of the executive at different times, we do not observe anything equivalent to government-versus-opposition voting.

FIGURE 7.

Scaling of the Northern Irish Assemblies: Government-versus-Opposition Pattern is Absent

Full key for the many parties is found in Appendix C.

In sum, the pattern of voting in the Welsh Assembly conforms to what we should expect given the unified control of the agenda by the governing majority. Removing this necessary condition for strategic opposition voting—that of unified agenda-setting power for the executive—leads to the converse prediction: the scaling of roll calls should no longer reflect a government-versus-opposition dynamic. Data from the Northern Ireland Assembly suggest the hypothesis is correct: in the Northern Ireland Assembly, with its “shared” executive, the government-versus-opposition pattern disappears. Thus, constitutional variation in the new assemblies introduced under the New Labour constitutional reforms can account for differences in the voting patterns of these assemblies. The evidence refutes the claim that government-versus-opposition voting is related to differences in policy domain (or that it arises as an “indigenous” aspect of politics in the British Isles).

IMPLICATIONS WITHIN AND BEYOND WESTMINSTER

Within parliamentary systems that provide strong agenda-setting powers to the government, the ability of the opposition to extract policy concessions via their voting strategies might vary according to the structure of the party system. We use the illustrative case of a single-party majority government with single-party opposition. Clearly, our findings extend directly to parliamentary situations where the government and/or opposition contains multiple parties and, in particular, to the case where the ideal points of government and opposition MPs lie, on average, at either side of a single cut point (that may, for example, distinguish a left voting bloc from a right voting bloc). If these conditions are not met, for example, where opposition parties can be found either side of the governing party, then we should still observe strategic opposition voting. Then, whereas it would not be possible for each opposition party to simultaneously extract policy concessions on the same policy dimension, each party would extract concessions on different bills. However, even in this case, if one party were the main party of opposition, and the status quo policies are, on average, closer to the ideal points of their MPs than to those of the government, then a strategy of outright opposition to the government will yield dividends to the minority.

Our focus has been on Westminster systems with their fusion of executive and legislative powers and majoritarian control of the agenda combined with strong party discipline. As shown by Hix and Noury (2008), government-versus-opposition voting is most prominent in Westminster-style systems but also observed elsewhere. It is possible then that our insights carry over to other parliamentary systems. Indeed, in many non-Westminster parliamentary systems, agenda setting is controlled by a minimal winning coalition and strong party discipline exists: thus, majoritarian control of the agenda should also give rise to strategic opposition. However, in non-Westminster systems, the role played by opposition voting strategies may be countered by other constitutional checks and balances; these issues remain ripe for further exploration.

CONCLUDING COMMENTS

Using new evidence from the first two Scottish Parliaments between the years 1999 and 2007, and comparing and contrasting it with evidence from the Westminster Parliament, we challenged the view that cohesive government-versus-opposition voting patterns can be explained by similarity of preferences within and between the government and opposition ranks. By exploiting the institutional variation between Westminster and Holyrood, we showed that government-versus-opposition voting can plausibly be linked with the key institutional similarities of these systems, namely, the procedural rules for reaching collective decisions in Westminster systems in which agenda-setting power is given to the majority party. Our formal analysis reveals that strategic opposition is central toward explaining the cohesive nature and distinct pattern of government-opposition voting. The opposition can gain by adopting “just say no” strategies that counteract the government's control of the agenda. Doing so reinforces the cohesion among government backbenchers, and the end result can be greater policy moderation. Our insights challenge conventional wisdom that the opposition is marginal to policy making in Westminster democracies.Footnote ²⁵