Intra-party decision making, party formation, and moderation in multiparty systems

Abstract

We establish coalitional stable party structures of a party formation game in an elected assembly. Farsighted political players can commit to form parties and to vote on policies according to the party position which is determined by intra-party majority rule. Parties may form governments and block proposals by a randomly selected member of the government. If the government recognition rule allows for the formation of multiparty governments, the median parliamentarian either realizes her ideal point or a policy lottery which she strictly prefers to the status quo. This outcome is enforced by the threat of forming a moderating centre party.

Introduction

This paper develops a model of endogenous party formation in an elected assembly in which parties form governments and governments select a policy by a random proposal maker process. In the last stage, the policy selection stage, a proposal is voted against the status quo. Each party coordinates its votes by intra-party majority rule but within a multiparty government, all parties have to accept the proposal. This is compatible with a view of parties taking a role in negotiating policy outcomes in multiparty governments and enforcing them in a parliamentary vote. At the same time each party is constrained by the need to win majority support among its own parliamentarians in the collective decision processes within the party. Seen from an ex ante perspective, with these assumptions a government generates a policy lottery. At the government formation stage, governments form for a given party structure. For the party formation stage, I establish stable party structures of a party formation game applying the largest consistent set (Chwe, 1994) as solution concept. I show that with a multiparty selection rule, under which a coalition of parties commanding an absolute majority is selected as government, the median parliamentarian realizes either her ideal point or a policy lottery which she strictly prefers to the status quo. Such an outcome is remarkable because the median parliamentarian is not pivotal in all governments which may form in stable party structures and may have to settle for worse outcomes than the status quo in some votes. Moreover, a multiparty government which preserves the status quo is conceivable, although such a government does not form in equilibrium unless the status quo coincides with the median voter’s ideal point. In the end, it is the option of forming a moderating centre party where the median voter is pivotal which allows us to eliminate all party structures where the median parliamentarian does not improve over the status quo.

I compare the multiparty selection rule to a majoritarian selection rule where a party with plurality in parliament is selected as government. This rule has effects which are comparable to multiparty selection provided that a parliamentarian majority is allowed to block proposals irrespective of party membership. Here, the paper relates to the debate, mainly regarding the US congress, on whether parties are effective in shaping policy outcomes and on the relative importance of the legislative median versus the median in the majority party in the legislative process.¹ In Europe, where multiparty government is common, parties are typically identified as unitary actors which negotiate political outcomes between them and can to some extent discipline the various factions of which they are composed (see e.g. Laver and Schofield, 1990).

Our results shed some light on empirical regularities found in different parliamentary systems. Multiparty systems are often characterized by small centrist parties (Schofield and Sened, 2006). Huber and Powell (1994) who distinguish the ideal forms of majority control and proportional influence systems–the latter being identified with more elements of power-sharing–find that government policies in proportional influence systems tend to be closer to the position of the median voter compared to majority control systems. This form of policy convergence is in spite of the fact that party positions in multiparty systems do not typically converge to the political centre (Schofield and Sened, 2006). Birchfeld and Crepaz (1998) find that systems of proportional representation (PR) are characterized by more redistribution relative to systems with plurality voting and relate their finding to the fact that under power sharing arrangements typical of PR systems the consent of more political players is required for accepting a policy proposal.²

This paper endogenizes the emergence of parties as veto players in an assembly under multiparty selection.³ What drives the formation of parties is the non-proportional relationship between positive agenda power, i.e. the power to set the agenda, and the size of the group formed which is typical of parliamentary institutions: for example, the biggest parliamentarian faction may be asked to form the government, a party or a coalition of parties may have to pass an inauguration with an overall majority to become the government,⁴ chairs in committees may go to the largest party. On the other hand, by joining a party agents may sacrifice negative agenda power, i.e. the power to block proposals, because they accept to vote with the party majority.⁵ Forming a small party increases negative agenda power provided that it is indispensable as a member in a government. In particular, in a small centrist party politicians are able to block more proposals than if they had organized themselves within a larger party. Here, this paper suggests a rationale for the formation of small centrist parties with ideologically motivated politicians. This is different from the conventional wisdom which suggests that small centre parties form because they enable office seeking politicians to increase the probability of entering the government.

Whilst we do not model the electoral stage and instead focus on the formation of parties in an elected assembly, the results still have obvious relevance for the formation of parties which form in order to win elections. It seems reasonable to assume that only such parties can credibly contest an election, which are stable against defections after the election is concluded.⁶

Only recently, a number of papers have begun to analyse the emergence of parties as coalitions of political agents. Austen-Smith (2000) combines party formation and policy selection. He models parties as representing endogenously formed economic classes and finds that policies with more redistribution are obtained under proportional representation but that policies under plurality rule are closer to the median’s preferences. Morelli (2004) determines effective parties and policies under proportional representation and majority voting. His setting involves multiple districts with a multitude of candidates running in each district. Bandyopadhyay and Oak (2004) present a citizen candidacy model in which groups of citizens select a candidate/party, and parties form governments. In Osborne and Tourky (2008) agents organize themselves in parties, committing themselves to vote for the same position. With economies of scale and with single-issue politics a two-party system emerges, irrespective of the electoral system. Kaminski (2006) focuses on the effect of electoral laws on party structure and hence on the stability of electoral rules and political parties. In his model parties are the movers which may select platforms, split or coalesce. If a coalition commands a majority in the chamber, it may change the electoral rule. Levy (2004) models platform selection where credible choices are restricted to the Pareto set of the party members. She shows that there are strategy choices which parties can adopt which cannot be matched by individual candidates.

There are a number of models where parties emerge as a response to an environment which creates uncertainty. In the coalition bargaining model of Jackson and Moselle (2002), legislators join parties to ensure that they are not excluded from a winning coalition. In Eguia, 2011a, Eguia, 2011b, legislators are uncertain about their true preferences but hold priors regarding their probability to vote for or against a proposal. In a stylized example, Eguia (2011a) shows that some players want to commit to vote with their party and form voting blocs whilst members in the middle of the spectrum refuse to commit. Similarly, in the present paper it is in the face of random proposal maker selection that parties allow legislators to realize preferred lotteries.

In a theory of party formation, parties have to change the set of equilibrium policy outcomes (Krehbiel, 1993). Eguia, 2011a, Eguia, 2011b, as does this paper, focuses on ex ante incentives for parliamentarians to commit their votes.⁷ In the bargaining model of Jackson/Moselle, a parliamentarian who has joined a party can be guaranteed her continuation payoff from reverting to a game without parties. The agenda setting power of parties and costly bargaining affect the policy outcome. As Diermeier and Vlaicu (2011) show, institutions which support unequal agenda setting power may arise in the equilibrium of an organizational game in parliament.

The present paper models parties as coalitions of political agents and describes the emergence of parties as an equilibrium phenomenon. But rather than purely focusing on party formation it explains, as Austen-Smith (2000) and, with a different focus, Kaminski (2006) do, party structures as they are characteristic of parliamentary democracies with proportional influence systems and identifies their common effect on policy outcomes. For this analysis, we require a solution concept which is sufficiently inclusive, i.e. which contains a rich class of potentially stable party structures and at the same time sets an intuitively convincing threshold for eliminating outcomes. In addition, we wish to employ a solution concept which achieves this end without imposing too many restrictions on agents’ moves, acknowledging that party formation potentially involves complex coalitional bargaining processes. The solution concept which we found most appealing given these criteria is the largest consistent set (Chwe, 1994).⁸ This solution does not preclude any coalitional move on which the agents involved unanimously agree. It does presume farsighted agents, i.e. agents which are deterred from a deviation which might result in an outcome which does not represent an improvement over their current situation. Thus, agents are assumed to behave conservatively, i.e. they have a predilection to the status quo situation. This assumption ensures existence of a solution and it specifies what we mean by a stable party structure.

In the end, all models of party formation have to deal with the issue of how to ensure stability in an environment, which naturally gives rise to moves by coalitions of agents. Levy (2004) also presumes farsightedness on the side of agents but applies equilibrium binding agreements as solution concept (Ray and Vohra, 1997). There, coalitional moves are restricted to deviations which result in finer, not coarser, coalition structures. Kaminski considers a myopic coalitional equilibrium which is proof against splits and mergers. Eguia (2011a) considers unilateral deviations and, in his application, split-proof equilibria, i.e. equilibria which are proof against further splits of parties when agents are myopic.

As we are interested in combining party formation and government formation, forward-looking behaviour is a natural assumption to make. At the party formation stage, farsightedness implies that parliamentarians take into account moves by other parliamentarians which are precipitated by their own move. This behavioural assumption is meaningful in a model of party formation. For example, the four Labour Party defectors which founded the UK Social Democratic Party, two of them sitting Members of Parliament, would have thought about which further members they would attract from the remaining Labour party (as they eventually did).

I set up the model in Section 2. Section 3 states the main results. Section 4 discusses the implications of our modelling decisions for our results.