Leaders and Labels: A Principal-Agent Perspective on Party

Position Taking∗

William B. HellerDepartment of Political Science

Binghamton Universitywheller@binghamton.edu

version 5.1

October 2015

Abstract

Who gets to make decisions inside parties? Political scientists have focused a great deal ofattention on relationships among political parties, asking for instance how parties’ legislativeand electoral contexts affect policy positioning and coalition bargaining, but what goes onwithin parties has most often been left to one side. How decisions are made and who is involvedin making them are fundamental questions, however, affecting not only how parties look tothe outside observer but also the kinds of decisions they make. In this paper, I explore therelationship between a party’s institutional context and its internal decision making. I derivetestable propositions as well as propositions that would in principle be testable were it feasibleto collect the requisite data.

∗Thanks to Seden Akcinaroglu, Marc Debus, Matthias Dilling, Javi Lorenzo, Shane Martin, Joshua Potter, BjørnErik Rasch, and participants in the Swedish Research Institute/University of Oslo–Uppsala University Workshop onThe Importance of Constitutions: Parliamentarism, Representation and Voting Rights for comments and advice, andto Diana Branduse and Allison Bugenis for cheerful, even if not always reimbursed, research assistance.

1 Introduction

Maurice Duverger (1976, viii) declared pessimistically in 1951 that “in another fifty years it might

be possible to describe how political parties actually function.” Fifty years have come and gone,

but Political Science’s understanding of intraparty politics remains limited largely to studies of

electoral incentives, the role and salience of factions, and individual cases. The focus is on the

what rather than the how of party behavior. Those who do look inside parties argue that parties

organize in order to achieve specific ends and to ensure that their elected agents work together

toward those ends. Party leaders are understood to be the captains of their organizations, wielding

the whip of discipline as necessary to ensure unity in pursuit of their collective goals.

This paper builds on the basic point that parties and their agents control the flow of parlia-

mentary work. I argue that a central element of party organization derives from the need to place

agents in positions of influence, so how parties organize depends on the extra-party positions (i.e.,

positions defined by legislative or constitutional, rather than party, rules) they can expect to con-

trol (cf. Carroll, Cox, and Pachon 2006; Cox 2006; Laver and Shepsle 2000). One key implication,

that titular party leaders typically cannot act unilaterally, stands in stark contrast to the usual

assumption that party decisions fall to the formal party leader (c.f., e.g., Kam et al. 2010). I assume

that party leaders and backbenchers alike value the clarity of their party label, but whether or to

what extent parties are unified is secondary to the question unified around what?

The content of decisions depends on who makes them. Moreover, the unity commonly ascribed

to carrots and sticks strategically wielded by party leaders (Laver 1999) and the party’s role as a

link between voters and the executive in parliamentary systems (Cox 1987; Epstein 1967; Laver

2006) turns out to be an endogenous consequence of how influence is allocated. The question of

what parties decide to do boils down to how decision-making influence is allocated to individuals

within them.

To make this argument, I construct a formal model that makes party positioning dependent on

the identity of party agents. I show that party agents and party position should be chosen jointly.

I arrive at this conclusion by building a model of party organization from three key assumptions.

First, individuals care about policy and prefer that their party’s position as well as policy outcomes

be closer to their own preferred policy than farther from it. Second, party leaders and rank and

1

file alike value a strong, coherent, clear party label. And third, party members have heterogeneous

preferences and beliefs about what policy positions best serve their own interests and those of the

party as a whole.

2 Parties in Context

From the outside, political parties are largely black boxes. Two groups stand out: party leaders

and their agents—ministers, spokespersons and rapporteurs whips, perhaps high-ranking members

of important committees—on one hand, and rank and file on the other hand. Scholarly views tend

to take the role of agency seriously, focusing on how leaders keep their agents in line and leaving

the rank and file as a backdrop; the concern is whether agents will toe the party line as well as

demonstrate minimal competence (cf. Dewan and Myatt 2010). In practice, studies concentrate not

on party leaders generally, but on prime ministers; and rather than the role of agency in intraparty

politics, they look at the interparty politics of coalition governments (see, e.g., Dewan and Hortala-

Vallve 2011; Laver 1999; Laver and Shepsle 1996; Saalfeld 2000; Strøm 2004; but see also Laver and

Shepsle 1990; 1999).

The literature, in short, treats publicly visible, apparently high-ranking party members as simple

agents of the formal party leader (operationalized most of the time as the Prime Minister). As such,

the working assumption is that they are motivated by the perks of office or other non-policy benefits

and not by policy goals. These non-policy ambitions give Prime Ministers leverage to manage and

discipline delegation to their ministerial agents (Dewan and Myatt 2010; Indridason and Kam 2008;

Kam and Indridason 2005). In coalition governments, however, ministers’ policy goals are treated

as paramount (e.g., Dewan and Hortala-Vallve 2011; Laver and Shepsle 1996; and cf. Osbourne

and Slivinski 1996). The upshot is that it shouldn’t matter for any given party who heads any

particular ministry, as long as all are reasonably competent. But accounts in the popular press

suggest otherwise, as do the actions of parties themselves. The existence of shadow governments is

evidence of this.

In some parliamentary systems, major opposition parties field front-bench “shadow ministers”

whose job it is to espouse their respective parties’ positions in contrast to those of the government.

Inasmuch as the key concern were ministerial competence and not preferences, the only reason to

2

expect much similarity between a party’s shadow government and its front bench in government

is a relatively small pool of ministrables (Paxman 2003, cited in Dewan and Myatt 2010, 268).

And even then the correspondence between any given agent’s position in shadow and government

ministries should be essentially random. But it is not.

What would random assignment look like? Consider shadow and government ministers for the

UK Labour and Conservative parties before and after the 1997 and 2010 parliamentary elections,

respectively. As can be seen in Figure 1, Tony Blair’s 1997 shadow cabinet numbered 21, including

Blair himself, while his post-election Cabinet was 20 strong. Eighteen individuals were in both, with

11 of those covering the same portfolio. The probability that at least one shadow minister would

end up in the analogous government portfolio is a reasonable 0.632; but the probability that eleven

or more would end up in the same position by chance is only slightly over one in 100 million. And

the probability that exactly eleven ministers would be randomly matched to their shadow positions

is less than one in 9 billion. The story in 2010 is similar: while only 18 out of 26 Conservative

shadow ministers made it into the Cabinet,1 12 of those held the same portfolio before and after

the election. As noted, this kind of correspondence is unlikely to happen by chance, which supports

the reasonable conclusion that there is more to being a minister than competence alone. Snapshots

of evidence from Australia and Germany reinforce this suggestion.2

0

5

10

15

20

25

30

Labour 1997 Conservative 2010

Shadow Cabinets, UK

Shadow Cabinet Both SameSeat

0

5

10

15

20

25

30

35

Labour 2007 Lib‐Nat 2013

Shadow Cabinets, Australia

Shadow Cabinet Both SameSeat

0

2

4

6

8

10

12

14

SPD 1971 (1965) CDU/CSU 1982(1976)

SPD 1998 CDU/CSU 2005

Shadow Ministers, Germany

Shadow Cabinet Both SameSeat

Figure 1: Cabinets and Shadow Cabinets in The UK, Australia, and Germany

Following the UK Cabinet since 2010 and through the May 2015 election reinforces the picture.

There were two major reshuffles, in September 2012 and July 2014, along with minor changes (e.g.,

1The limited number of Conservatives who were in both the shadow cabinet and the Government should notbe surprising given the circumstances. First, there were fewer Government than shadow portfolios; and second theLib-Dems took four portfolios for themselves.

2The number of German shadow ministers who move into Government is quite low, except for the SPD in 1998,but note that for the 1971 and 1982 Governments the available shadow-minister data are from some six years (i.e.,one full government term) earlier.

3

brought on by scandal or impending retirements) along the way. Looking only at the Conservative

party, those changes brought 16 MPs and two members of House of Lords into the Cabinet who had

not been there from the beginning of Cameron’s Government in 2010; of those 18 “new” members

of the Cabinet, 11 continued on in the new, single-party Conservative Government after the 2015

election. Cameron’s second Government comprised 24 Cabinet portfolios and included only four

MPs who had not served in some capacity in the previous Government.

The record from shadow governments and reshuffles suggests that while the list of those consid-

ered able to run a ministry might be fairly short, it does leave room for choice. Ability might be a

strong criterion for inclusion on a party’s front bench, but it is not the only criterion. New people

enter—and are moved around in—government over time, and at the end of the day it is clear that

there are more ministrables than there are available ministries, albeit not all that many more. If

these positions are valuable, party members should be willing to pay some costs to get into them;

but if they are not valuable, then why would anyone undertake to fill them? In short, what do

shadow ministers, spokespersons, and ministers get out of their positions? And how does who they

are (and what they want) matter?

Studies that seek to “lift the lid” on intraparty politics (Laver 1998, 22) tend to focus on specific

cases and interfactional competition (see, e.g., Cox, Rosenbluth, and Thies 2000; Di Palma 1977;

Giannetti and Laver 2005; Mershon 2001a; b; Panebianco 1988; see also Amorim Neto and Santos

2001; Belloni 1978; Bettcher 2005; Carty 2004; Cyr 1978; Fukui 1978; Green-Pedersen and van

Kersbergen 2002; Ignazi and Ysmal 1998; McCubbins and Thies 1997; Nicholson 1978; Roback and

James 1978; Sabatini 2003; Zuckerman 1979). Throughout, electoral incentives loom large (e.g.,

Cox, Rosenbluth, and Thies 1999; Moreno 2005; Morgenstern 2001; Samuels and Shugart 2010);

where they do not, coalitions and coalition bargaining take their place (e.g., Laver and Shepsle

1990; 1996; Mershon 2001a; b). How the individuals who populate parties engage in collective

action and move their parties to action largely is treated as beside the point.

That the details of how parties organize remain hidden is consistent with Duverger’s (1976)

lament. The appearance of party agency and the principal-agent theory that underpins the entire

concept of democratic representation jointly yield a puzzle, however. Parties exist in large part to

exercise power, to which end they seek to control offices. But the offices in question can be held,

and their authority wielded, only by individuals; and individuals are imperfect agents. How, then,

4

to reconcile the image of unified parties with parties that actually get things done?

To address this question, I treat parties not as aggregators of preferences but as creatures of

individuals’ ambitions (cf. Aldrich 1995). The latter implies the former: party “preferences” are

defined by influential party members’ individual preferences. A given member’s influence is created,

shaped, and constrained by his party’s aspirations to legislative influence and how it organizes to

exercise the same. The electoral connection is vital, and in what follows I assume that a strong

electoral connection depends on a strong party label, which in turn requires party members to back

each other up in word and deed.3 Party members, rank and file and leaders alike, therefore hold

party unity particularly dear—enough to structure party decision making with party unity in mind,

but not enough to overcome the dilemma embedded in the logic of collective action (Olson 1965)

where no individual can expect to do enough harm to the label to make it worthwhile to set aside

his own interests in favor of the party’s.

The linkage between legislative structure and party organization echoes Cox and McCubbins’s

(1993) sketch of parties as legislative cartels. Parties give agents jurisdiction over defined policy

areas and let them make policy via an intraparty logroll. Of course, agents who collectively define

the party position essentially are making decisions for the party, something I take to be a privilege

reserved for leaders. Policy authority thus implies leadership rank. In the US Congress, this means

that committee chairs (and, for the minority party, ranking members of committees) effectively are

leaders. To extend the argument to parliamentary systems I focus on government ministries.

A given party can reasonably hope to control a more-or-less predictable number of extra-party

positions (broadly, “mega-seats”; see Carroll, Cox, and Pachon 2006) in the aftermath of elections

and coalition bargaining (Laver and Shepsle 2000). If the individuals who fill those posts can affect

what the party does or how it is perceived, they also are in a position to hurt the party if their

behavior is at odds with its established position. And if they both care about policy and have

preferences different from the party line, they cannot commit to do otherwise. The question then

becomes whether parties can organize so that what their agents say and do align with their labels

or, instead, they leave to their formal leaders both the benefit of setting party policy and the costs

of maintaining discipline.

To get at this question I formally define a strategic context in which individuals seek to maximize

3here briefly highlight literature on value of blurred party labels

5

their own utility as members of a party. What happens inside the party might not affect legislative

outcomes—a small opposition party is unlikely to be able to influence much—but an actor who has

no influence in her party certainly will have no influence in the legislative process more generally. I

show that the more important is a clear party label, the more likely it is that party decision makers

and publicly visible party representatives will be the same people; and the decisions they make

about party positions will balance their own preferences and beliefs (e.g., about voter preferences)

with their desire to maintain party-label clarity. Along the way I find, surprisingly and contrary to

conventional wisdom, that in equilibrium the costs typically thought to keep party agents in line

are irrelevant.

Throughout this paper, “party members” refers (interchangeably with “legislators” or “repre-

sentatives” unless context suggests otherwise) to individuals who have been elected or appointed

to office—I focus on those who occupy legislative seats—under the mantle of some party. Activists

and supporters are important, of course, but lie outside the scope of my analysis.

3 Parties and Party Members

A political party is defined by its label. The label comprises perceptions of the party with respect to

such considerations as its policy positions, member honesty and competence, the constituencies it

seeks to serve, its prospects for policy influence, and its reliability. Some of these things, chiefly party

position, party members can affect; the rest are “valence” issues (cf., e.g., Adams and Merrill 2008;

Ansolabehere and Snyder 2000; Groseclose 2001; Schofield 2007; Schofield and Sened 2005; 2006),

which cannot be manipulated intentionally. As long as leaders believe that voters care where parties

stand on policy they will seek to ensure as clear a label as possible.4

To begin, a policy position x ∈ Rm = (x1, x2, . . . , xm) is a vector of positions in m-dimensional

policy space. A party policy position is a vector xp = (xp1, xp2, . . . , xpm). If the status quo is

xq = (xq1, xq2, . . . , x

qm), then the position of a party that takes no position at all on the ith and jth

dimensions is xp = (xp1, xp2, . . . , xqi , x

qj , . . . , xpm). That is, for any dimension on which the party

takes no position it effectively adopts the status quo position on that dimension as its own.

4Party leaders (some elected legislators, some not) and sitting or former legislators I interviewed in Denmark,Finland, The Netherlands, Spain, and Sweden in summer 2009 all underscored the importance of the party label inelections as well as legislative bargaining.

6

Taking position xpi on the ith dimension requires sending a signal. The intended recipient of

the signal is irrelevant for present purposes, but in the real world the signal probably would be

intended for voters and, possibly, other parties’ leaders. If a party sends different signals about its

position on at least one dimension, its perceived position is no longer a specific point, but rather a

distribution, call it Ω; given two competing signals, x1p and x2p, of party position xp, for example,

xp ∼ Ωx1p,x

2p

(see Figure 2 in Section 4.1).

The problem with identifying a clear party position (for which no signals on any dimension

conflict) is twofold. First, arriving at a position is subject to all the social-choice instability problems

that confront any collective decision-making endeavor. And second, once a position is established

it is not obvious that anyone with authority to implement it or to communicate it to others would

do so accurately. I elaborate on these problems below.

If party decisions are taken collectively, setting a consistent party position in day-to-day politics

should be problematic. There is no a priori equilibrium position (though see Plott 1967), and absent

restrictions on individuals’ participation in decision making no decision (or at least no coherent

decision) is likely (cf. Cox 2006). Such restrictions imply structure and process.

A political party P incorporates individual members l ∈ P with ideal points xl ∈ Rm. Party

size does not matter as long as it can support a sufficiently large leadership group (Laver and

Shepsle 2000; see Corollary 2). I do not constrain party-member preferences to be similar, much

less identical, on any dimension. Members’ revealed preferences are constrained, e.g., by constituent

pressures or by decisions taken in party congresses, but those pressures define the party’s feasible

policy space Xp ⊆ Rm rather than any specific point. Within that space, each member’s preferred

position is defined by the combination of his own policy preferences and beliefs about the kinds of

position that voters and activists will reward. That members have their own ideal points in policy

space means that they are not simply automatons driven by reelection concerns or rent seeking;

each prefers the party’s position to be close to his ideal point, all else equal.

What might a party agent gain from signaling something different from the established party

position? If the agent has policy influence, the answer is obvious: she gets policy outcomes closer to

her preferred policy. If she is in opposition, by contrast, she can affect perceptions of what her party

stands for, but not policy outcomes. If a party prospers despite having a muddied label, however,

leaders might seek to regain label clarity by moving xL closer to the defecting agent’s signal;

7

this effectively increases the agent’s influence in the party and, should the party gain legislative

influence, over policy outcomes.5

The standard take on how parties communicate clear positions is that, at least in parliamentary

systems, the efficient secret (Cox 1987) and parties’ value as vehicles for advancing political careers

(Laver 2006, 126) rein in the potential for dissent. This answer sidesteps the crucial question of

what the party position (around which its members unify) is. The official party position xL is a

product of intraparty decision-making procedures. Inasmuch as party decisions are the purview of

a privileged leadership cadre (cf. Michels 1915), any l ∈ P able to influence xL is by definition a

member of that august group. In day-to-day legislative politics, the party position is communicated

(translated into legislative behavior and public statements) by party agents—individuals who are

generally understood to be speaking for the party. If the position communicated by agents is

different from the public, leadership-established position, the result is corrosion of the party label.

Standard approaches to parties either ignore the problem or generate more questions than

they answer. In the first instance, assuming that parties are unitary actors might make sense for

analyses interested in party behavior, but ought not be used as an excuse to treat intraparty politics

as trivial. Similarly, to accept heterogeneous preferences among copartisans but then assume that

they all value the party label (and the party unity that underpins it) enough to toe the party line

begs the question because it is tantamount to assuming that politicians value party unity more than

party policy, which in turn at the very least raises the issue of whether party unity is instrumental

or an end in itself. With respect to generating questions, approaches that rely on the threat of

discipline to keep agents in line—where the party’s positions begin and end with the formal leader,

who wields carrots and sticks to ensure that agents faithfully toe her party line—are problematic

for two reasons. On one hand, they require the leader to spend time and resources monitoring (and

possibly punishing) agents. Presumably, that is costly. On the other hand, it is difficult to imagine

that there is much incentive to be an agent if the job implies both serving as a target for public

ire when policies strike a sour chord and an inability to influence what those policies are.6 Instead,

5refer back to note 26For backbenchers, the need for discipline should essentially be nil (see Hu 2013; Hu and Heller 2011) as long as

advancement is valuable and the probability of advancement is sufficiently high—which is more likely if membersare promoted into leadership fairly regularly, as appears to have been the case during Cameron’s first Government.The logic is as follows: Party leaders promote backbenchers to leadership. Leaders care about policy and so preferto promote backbenchers who share their preferences. Consequently, ambitious backbenchers will evince agreementwith current leaders. Once given influence in party decision making, those whose loyalty was insincere can move the

8

I argue here that there ought instead to be a close relationship between who gets to represent the

party in publicly visible positions and who makes decisions for the party (cf. Samuels and Shugart

2010 with respect to parties in presidential systems).

4 The Logic of Party Positioning in Institutional Context

Two conditions must hold for a party position to be supported by its own members’ signals about

its location. First, the position must be an equilibrium. Given the generic instability of mul-

tidimensional collective choices (e.g., McKelvey 1976; Plott 1967; Schofield 1983), this implies a

structure-induced equilibrium (Shepsle 1979). If there is only one decision maker, this is trivial.

With two or more decision makers, however, things get more complicated. How and why depends

on the larger institutional context and the party’s position within it.

The second condition is that no one with authority to send signals about the party should do

so inconsistent with the leadership-established position. Since no agent can commit to signal a

position different from her ideal point (cf. Osbourne and Slivinski 1996; Laver and Shepsle 1996),

that directly implies that agent ideal points should match or closely track the party’s position. In

order to ensure that agents will faithfully play their roles, therefore, party positions must be set

with agent preferences in mind.

The point is that the establishment of a party position and the task of communicating that

position beyond the confines of the party are linked. The crux of the link is the relationship between

internal party organization and the selection of party agents. To analyze this relationship I model

a sequence of actions and decisions by party leaders and their agents and then back out a basic

logic of party organization from that. Throughout, I assume rational actors with heterogeneous

preferences.

4.1 Making and communicating party decisions

All actions are the purview of members of party P and directed toward defining the party position

xp. Each party member l ∈ P (with ideal point xl) can be part of party leadership L ⊆ P , one of

a set A ⊆ P of publicly visible party agents, neither (l ∈ P \ A ∪ L), or both (l ∈ A ∩ L). That

party position more in line with their own preferences.

9

Table 1: Summary of Notation

P A political party with n members l ∈ Pxp Party P ’s perceived position in m-dimensional policy space Rm

L The leadership of party P , L ⊆ PxL Party P ’s policy position as defined by l ∈ LA Publicly visible members of party P , A ⊆ Pli Agent l ∈ A with responsibility for dimension ixlp A signal about the party position sent by each l ∈ Acp Cost paid by each member of party P when p(xp = xL) < 1clA Cost imposed on party agents by leaders when xlp 6= xLcL Cost paid by leaders (each l ∈ L) when p(xp = xL) < 1di Basis vector for the ith dimension in Rm

α A switch indicating whether an agent is a member of party leadership, α = 1 if l ∈A ∩ L

ΩxL,xlp:l∈A Distribution of beliefs about party position xp after agents send their signals

is, leadership and agent roles are not mutually exclusive. All players have symmetric, unweighted

Euclidean preferences. When the game begins there is a leadership cadre in place—it could be a

single individual, or it could be several—chosen by Nature. It is the job of these leaders to choose

both a party position xL = (xL1, xL2, . . . , xLm) ∈ Xp and the party agents l ∈ A tasked with

communicating the on-paper label to the world through word and deed.

Party agents, publicly visible and understood to be speaking for the party, might be involved in

intraparty decision making. Alternatively, their job could be to uphold and promote their parties’

and policies in the give and take of legislative politics. In opposition or in government, each l ∈ A

carries out her duties by signaling xlp ∈ Xp; if the agent’s remit is limited to a subset of dimensions

in Rm, then she cannot signal a party position on a dimension beyond her authority. (Table 1

provides a handy reference to the notation.)

I limit analysis to cases where each party agent has authority over only one dimension, following

Shepsle 1979, and no two agents share authority over the same dimension. I assume further that

the number of available portfolios is exogenous (but see Mershon 2002).

Leadership does only three things in this game—they decide on their preferred party position,

xL ∈ Xp, they choose agents l ∈ A to represent the party to the outside world via signals xlp ∈ Xp,

and they promote rank-and-file members to leadership as necessary. I do not model this third task.

Note that while how xL is chosen can be opaque, communication of the position is not. For all

10

intents and purposes, it as if once leaders have decided xL they write it down and post it in public

view.

In sum, the game is confined to intraparty politics: all players are members of a single party

and hold whatever authority they have by virtue of their party affiliation. The game begins when

party leaders l ∈ L ⊆ P decide on a party policy position xL ∈ Xp ⊆ Rm and choose party agents

l ∈ A ⊆ P to represent the party to the outside world. Once selected, each agent li ∈ A sends a

signal xlpi ∈ Xp of party P ’s dimension-i position. The combination of the public leadership position

plus all agent signals determines xp ∼ ΩxL,xlpi:li∈A

, where if xlpi = xLi∀l ∈ A then xp = xL. Figure

2 illustrates a party position as perceived by voters conditional on leadership and agent signals,

with a single agent. When agent and leadership signals match, as in Figure 2a, voter perceptions of

the party position (Ω) center on the established position.7 As signals diverge, as in figures 2b–2d,

voter perceptions are tainted by confusion over which signal to believe and how to interpret each

signal in light of the other. In effect, each signal pulls perceptions of the other toward it so that

both are skewed and Ω is both spread out and fairly flat. As signals diverge, in other words, voters

lose track of what the party stands for.

Where will the party position be? Recall that any given agent’s ideal policy position is a

function both of the agent’s sincere preferences and her beliefs about how best to achieve policy

outcomes as close to those preferences as possible. The latter implies that agent behavior is geared

toward attracting voters. For each agent, the best outcome is a party position that is close to her

ideal point, at least probabilistically. For party leaders, the best outcome is to find some xL that

maximizes their collective payoffs. To do so they have to minimize any agent’s incentives to signal

a position different from xL. Given the near certainty of agency loss, it is not immediately obvious

how they can do so. To answer the question, therefore, I begin with some agent’s signal and work

backward.

4.2 Principals, agents, and party positions

There are two points to keep in mind. First, how close together agents’ preferences are should

depend in part on such exogenous factors as the party system, ease of party switching (Heller and

Mershon 2009), electoral rules, or the diversity of the electorate and the degree to which any given

7The figure incorporates an assumption that voter perceptions of all signals are noisy but unbiased.

11

(a)

W HPerceived Party PositionLPerceived Leadership Position

Perceived Agent Signal

xpl

xL

(b)

W

xL xpl

(c)

W

xL xpl

(d)

W

xL xpl

Figure 2: Effect of Agent Signal on Ω

candidate has to cater to a specific segment thereof, along with historical and sociological factors

beyond the scope of this paper. Second, politicians’ ideal points are a product of some combination

of their ‘raw’ or ‘true’ preferences (untinged by strategic or practical considerations) and other

concerns dictated by ambition, career, and expectations or beliefs about what is feasible. I assume

also that actors have well-behaved spatial preferences over policy positions.

Typically, agents are seen as constrained by party discipline—essentially a set of costs. Hence,

in principle, the signal that each party agent sends should be constrained, as all members pay

cp when those signals diverge from the leadership position. For any agent who also is a party

leader (i.e., l ∈ A ∩ L), divergence between xL and xp imposes the additional cost cL. This is

functionally equivalent to making leaders desire a clear party label. From the agent’s perspective,

however, the effect of her actions on cp and cL are marginal. That is, her signal cannot depend on

12

her expectations about her counterparts’ actions; and no agent can hope that her signal will affect

what any other agent does.8 Defining α ∈ 0, 1 as a switch that indicates whether an agent also

is a member of party leadership (l ∈ A ∩ L −→ (α = 1)), the optimal signal for agent l ∈ A thus is

xl∗pi =

xL if ul(xL) ≥ ul(xp ∼ ΩxL,x

lpi 6=XLi

)− (cp + clA + αcL);

argmaxxlpi 6=XLi

ul(xp ∼ ΩxL,xlpi 6=XLi

)− (cp + clA + αcL) otherwise.

Rearranging terms, any agent l ∈ A will signal xL only if

ul(xL)− ul(ΩxL,xlpi 6=XLi

) ≥ −(cp + clA + αcL). (4.1)

If the agent signals a party position closer than xL to her ideal point, the left side of Equation

4.1 necessarily will be negative. Consequently, while the right side of the equation also always

is negative, the inequality need not hold. The farther is xL from l’s most preferred point in Xp,

the less likely is the agent to toe the party line, all else equal. Thus, holding costs constant (and

understanding that party leaders prefer to keep cL low), the problem for the leadership is to select

agents with ideal points on the pertinent dimensions at or close to xL.

It is useful at this point to define agents’ jurisdictional authority. Let di define the ith dimension

in Rm, so that D = di : i ∈ Rm is an orthogonal basis for Rm. An agent l’s jurisdictional authority

is the set of dimensions on which she can send meaningful signals about her party’s policy position.

For the formal analysis, I allow agents authority over only a single dimension. Define dli as l’s

jurisdictional authority (cf. Shepsle 1979, 32), so l can signal only xlpi; if xlpi = xLi, her signal

reinforces the perception that xp = xL. If, by contrast, she signals xlpi 6= xLi, she weakens that

perception, essentially skewing ΩxL,xlp:l∈A toward xlp and reducing its density around xL.

Given a leadership-established party position, the optimal choice CL(lA) for each dimension is

an agent who will toe the party line. Following Equation 4.1, that implies that any benefit the

agent might gain from signaling something different from the party position must be outweighed

by the costs she would pay for so doing. Thus,

8Detailed discussion of the costs—which, as noted, turn out not to matter—is in Appendix B.

13

C∗L(lA : ul(ΩxL,xlp 6=xL

)− ul(xL) ≥ cp|xlp

+ clA + αcL).

Taking xlp as l’s induced ideal point in X, this is tantamount to choosing agents such that xli = xLi.

In order to mitigate agency loss (as a consequence of agents sending signals xlp 6= xL), leaders

have to choose xL strategically so that each party agent can commit to signal xlpi = xLi. The

alternative is to search for members whose preferences match xL on some dimension, which would

introduce potentially prohibitive transaction costs. To that end, they must select agents from

within leadership or choose agents from among backbenchers and then define xL around their

preferences—which is equivalent to letting the selected agents qua leaders set party policy.

Leadership decision making. In order to exercise their influence, which they do in the first instance

by setting xL, the party leadership first has to discover a set of acceptable positions and then define

a unique policy package as the leadership position.

Definition 1. Define the leadership Pareto set xL as the set of Pareto optimal policy packages

for all members of leadership, xL = x ∈ X : @y ∈ Rm s.t. ul(y) ≥ ul(x)∀l ∈ L. Further define

CL(lAi) as the leadership choice of agent to represent the party position on dimension i.

Proposition 1. The optimal choice of agent for dimension i is the member of leadership whose

ideal point defines the party position on that dimension. Formally, C∗L(lAi) = l ∈ L : xli = xLi.

(Proof in Appendix A)

The advantage to being a member of party leadership lies in being able to define the feasible

set from which the party position is drawn. Leaders whose ideal points are used to define the party

position also benefit from having their preferences reflected in the party position. Backbenchers, by

contrast (consistent with conventional wisdom), have no influence in party decision making. The

only reason for a backbencher to toe the party line is general agreement with the party position or

the expectation of being able to affect that position in the future. Backbenchers who find themselves

often at odds with their party’s position and who have little expectation of being promoted to

influence might do better if they could find a new party home (cf. Heller and Mershon 2008; 2009).

While the conditions that affect backbenchers’ ambitions (and the availability of outside options)

will vary across systems, two interesting implications flow directly from Proposition 1 and the facts

14

that a) agents-as-leaders reveal their preferences sincerely and b) no backbencher who would like

the party position to be different has any incentive to say so (Lemmas 1 and 2 in Appendix A).

Corollary 1. The costs related to agent signals that are different from the leadership-defined party

position are irrelevant.

Corollary 1 provides a counterintuitive spin on the conventional understanding of discipline,

particularly as expressed through collective Cabinet responsibility. Party agents, including espe-

cially Cabinet members, toe the party line in equilibrium not because of the threat of punishment

but because it is they who define it from the outset. Costs matter only in the background as a con-

sequence of the assumption that the players hold party-label clarity to be important. Leaders don’t

pick backbenchers to be agents because doing so puts the party label at risk, and, consequently

should never have to hold out the threat of, much less apply, punishment for sending discordant

signals.

What about coalition (as opposed to party) discipline? While this issue is beyond the scope of

the present discussion, it bears reiterating that the similarity between the model laid out here and

the Laver and Shepsle (1996) Portfolio Allocation Model of coalition formation and policy making

is not coincidental. Both are based on Shepsle’s (1979) Structure-Induced Equilibrium and build

on the notion that the party or coalition agents who matter can commit neither to push nor defend

policy different from their ideal policy. Collective responsibility follows directly, as long as ministers

stick to their jurisdictions.

Corollary 2. Party leadership should comprise enough members to fill all available agent positions.

Proposition 1 establishes that where a party has to delegate to agents it chooses them from

within the party leadership, where the leadership comprises all party members with influence in

establishing the party position. Any given leader’s influence on any given policy dimension might

be hard to detect, but it is greater by leaps and bounds than backbenchers’ influence. It follows

straightforwardly that the party must include in its decision-making leadership at least enough

members to occupy any positions that the party reasonably can hope to fill. Parties that expect to

be in Government thus should have more expansive leaderships than parties that do not.

15

5 Discussion

The preceding model’s primary implication is that anyone who is or might be seen to be speaking

or acting on behalf of a party should be involved in that party’s decision making. At the level of

party backbenchers, whose legislative votes reflect on their party, this is incorporated in the usual

assumption that leaders have to keep backbenchers happy if they want to retain their jobs. This

lends itself to complex and nuanced empirical exploration: Where committees are important (see

Strøm 1990), for example, can a party’s ranking committee members be distinguished from party

leaders? (I would argue that, at least in the case of the United States, they cannot.) Or how

do parties that control significant subnational authority, e.g., in provincial or city governments,

balance the need for a clear party position against the natural tensions that arise between national

and subnational governments?.

These are questions worthy of further exploration. Here, however, I focus on a more obvious

conjecture: parties with a realistic chance at governing should field a pool of credible spokespersons

while they are in opposition. When their party gains office, those spokespersons should slide directly

into government positions. Indeed, they should move into positions that mirror the jurisdictions

they covered while in opposition, unless moving into government also requires interparty bargaining,

in which case a clear one-to-one correspondence between title in opposition and government portfolio

is less likely. Under no circumstances, however, should unknown backbenchers move into positions

of influence or high visibility. There should be, in other words, a close correspondence between

those who speak for the party while it is in opposition and those who comprise its cabinet cohort

in government.

One interpretation of the story here is that ministers get to where they are not so much because

they have the good fortune to become leaders and then hold the “right” position on some policy

dimension, but because they represent important interests within the party. In this view, Iain

Duncan Smith made it into the 2010 Conservative–Lib-Dem cabinet because of his associations,

not his positions.9 Following the reasoning through, at least some ministers are in the cabinet in

order to salve rebellious or otherwise contrary interests in the party.

It is more than plausible that some ministers serve as envoys from important groups in the

9I am grateful to Jane Green for pointing out this alternative interpretation.

16

party. However, it strikes me as unreasonable to suppose that they should consequently exercise

little authority, never mind having less authority than other ministers. If they did they would be

figureheads only, and it is hard to imagine that their supporters would suffer such window-dressing

representation quietly for long. Rather, their backers should expect them to carry out their duties

to benefit their supporters (and themselves), just as I argue above. Indeed, the only gap between

this “Iain Duncan Smith” story and my argument is that in the story the relevant positions on a

specific subset of policy dimensions are known ahead of time. But are they, and does it matter

if they are? The answer is no. If the group in question essentially has property rights over the

relevant dimension, then it is the group and not the party leadership as a whole, that sets policy

on those dimensions. The group then has to decide on who will represent it in cabinet, in which

case its best bet is to replicate the situation depicted in the lemma.

6 Conclusion

This paper argues that the allocation of influence in parties should reflect their institutional con-

text, more so for parties that reasonably can hope to wield policy influence. The nature of political

institutions that concentrate a party’s collective influence in the authority wielded by individuals in

specific offices implies that a formal party leader cannot realistically define party positions unilat-

erally. Instead, the agents in question define the party line, which they then represent faithfully by

following their own preferences. This holds for Westminster-system parties as well as continental

ones. The immediate implication is that analyses that restrict party leadership to such intra-party

offices as party leader and party whip might both overstate the significance of those offices and miss

important aspects of intra- and interparty politics alike. One such aspect, for instance, is that the

carrots and sticks commonly thought to induce party discipline do not matter in equilibrium; the

reason that they do not matter is not because they are unobserved (after all, a punishment that is

never used because it successfully deters punishable behavior clearly matters), but because the fact

that the choice of party position is endogenous to the choice of party agents makes them irrelevant.

This is not to say that formal leaders do not matter—at the very least, they play an important

role in selecting the party agents who define its label—only that their role, detailed examination of

which is beyond the scope of this analysis, is different from what the conventional wisdom suggests.

17

The conventional wisdom is that agents toe the line because they might lose their influence if

they do not, but this is unsatisfactory. Not only should the threat of punishment for deviating from

the party line make holding party-controlled mega-seats fairly unattractive for policy-motivated

legislators (since it would imply that a party agent gains no policy influence), but it demands as

well a level of resources spent on monitoring agent behavior that could more fruitfully be put to

other uses. Moreover, punishing defecting agents might suggest to voters that leaders are too willing

to employ harsh methods to maintain their influence (Huber 1996). My argument, by contrast,

suggests that spreading influence over party positions among a larger leadership group and then

drawing from that same group for party agents is a more efficient solution. What an erstwhile

sole leader loses in terms of policy position by sharing responsibility for party policy positioning is

compensated by both the peace of mind that comes from knowing that agents have zero incentive

to shirk and the liberty to put to better use resources that otherwise would have been wasted on

discipline.

As is often the case in delving into interesting problems, the answers I have found suggest more

questions worth exploring. Four in particular clamor for attention. Foremost among these is the role

of the titular party leader. The logic that yields Corollary 2 builds on the notion that leadership is

defined by decision-making influence, not formal titles. Parties do have named leaders, however, and

it would be rash to claim that they have no influence. What, then, do they do? Further, Corollary

2 places a lower limit on the size of party leadership, but suggests no upper bound. Should an upper

bound exist, or could a party (and its decision-making members) gain from extending a share of

influence to every member, as could be argued was the case for the German Greens before Joschka

Fischer’s Realos took control? Third, does the logic of the argument work only where single-party

majority governments are the norm, or can it be adapted to coalition situations? And finally, given

that the formal model allows a given individual to speak on only one policy dimension, would the

results still hold if a single individual were allowed to speak for the party on multiple dimensions?

The model here can provide leverage for delving into these and other issues, but definitive answers

will have to wait for future work.

This paper is about legislative parties. The intuition underpinning it is far broader, however.

Basically, individuals organize with others to achieve things that they cannot achieve working alone.

They organize out of self-interest. This implies that how organizations are structured should de-

18

pend less on what people are trying achieve than on the nature of the obstacles that make unilateral

action ineffective. Party organizations should look different in presidential versus parliamentary

systems, for example (cf. Samuels and Shugart 2010), or in bicameral versus unicameral legislatures

(VanDusky-Allen and Heller 2014). Similarly, interest groups should organize (and behave) differ-

ently in different systems. And, farther from the political arena, the perspective here can yield

insight in such varied areas as corporate strategies in response to accounting (or trade, or tax, or

employment) rules or how sports teams organize themselves and their play on the field. The key

problem, as in all collective action, is that while individuals need the group to get what they want,

what they want might well not be what is good for the group.

19

Appendix A Proof of Proposition 1

To prove Proposition 1, I begin with two foundational considerations. First, as indicated in the

choice of l∗ ∈ A, agents are selected on the basis of their preferences vis-a-vis the party’s established

position. Second, xL is defined as the dimension-by-dimension aggregation of leadership ideal points

on each dimension in X. In essence, members recruited into leadership are assigned influence over

a specific dimension.

Lemma 1. Leaders sincerely reveal their ideal points on the dimensions they influence.

Proof of Lemma 1. The proof is simple. Let xk be the ideal point of some (k ∈ L) with influ-

ence on the ith dimension. Suppose that k sets xLi = xk′i, xk′i 6= xki, so that (by definition)

uk(xLi) ≤ uk(xki). Then xL = xL1, xL2, . . . , xk′i, . . . , xLn. But then uk(xL) < uk(xL′), where

xL′ = xL1, xL2, . . . , xki, . . . , xLn, so k would have been better off setting xLi = xki.

By Lemma 1, xL must lie inside or on the leadership Pareto hull, defined on each policy

dimension by the ideal point of the leadership member (l ∈ L) with authority over that dimension.

The result is a unique point. Thus far, Proposition 1 follows almost trivially. It remains to be

established whether the opportunity to be a party agent could induce backbenchers to reveal their

preferences sincerely. It does not.

Lemma 2. Backbenchers who would like to change the party position have no incentive to reveal

their preferenes sincerely.

Proof of Lemma 2. Backbenchers play no part in defining the party leadership position, xL. There-

fore, the advantage that accrues to a backbencher who is chosen as an agent comes from being able

to signal that xL is closer to her ideal point than it actually is, i.e., xlp 6= xL. The only back-

benchers who would send such a sincere signal are those for whom the inequality in Equation 4.1

does not hold, so that ul(ΩxL,xlpi 6=XLi

) − ul(xL) > cp + clA + αcL. Because the signal they would

send would damage the party label, irrespective of costs, party leaders would never select them

to be agents if they were to reveal their preferences sincerely. Other backbenchers, for whom the

relevant inequality does hold, gain nothing from being named agents and hence have no incentive

to reveal their sincere preferences as long as doing so damages their prospects for entering the ranks

20

of leadership. Further, because even in the absence of costs leaders have a dominant strategy to

appoint agents whose preferences they know, only backbenchers whose preferences align with the

party preferences established by xL have an incentive to reveal their preferences sincerely, but in

doing so they contribute to a pooling equilibrium in which they are indistinguishable from all other

backbenchers whose behavior is strategic.

Proof of Proposition 1. Let clA∗

be agent cost on top of cp that induces an agent not in party

leadership to faithfully signal the party position. Suppose party leaders choose just such an agent

l ∈ P \ L to represent the party on the ith dimension. By assumption, ∃l ∈ P \ L : clA∗> clA, so

E(xp ∼ ΩxL,xlp) 6= xL. But if leaders choose agents from among themselves so that C∗L(lAi) = k ∈

L : xkpi = xLi∀i ∈ Rm, then xp = xL.

Appendix B Defining Costs

The primary source of costs is linked to the foundational assumption that party-label clarity is

valuable: inasmuch as signals different from xL reduce the clarity of the label, they thus im-

pose costs on all party members. The first of these costs, borne equally by all party members,

is cp = f(p(xp = xL), s(ΩxL,xlp:l∈A)), where s(•) is a measure of the distribution’s spread and

∂cp∂p(xp=xL)

< 0 and ∂cL∂s(•) > 0. I assume that this cost is subject to the classic collective dilemma,

whereby the damage cp does to any individual agent’s utility is less than her benefit from moving

the party position closer to her own. If the cost were greater, no agent would ever signal anything

but xp = xL, but then there would be nothing to gain from being an agent.

The second source of costs lies in the possibility that leaders can punish agents whose signals dis-

agree with the party’s leader-established position. This cost, cAl = f(‖xlp − xL‖),∂cAl

∂‖xlp−xL‖

> 0,

could for example be imposed through a Cabinet reshuffle or outright demotion, or in reduced levels

of party support for a given candidate at election time.

Finally, party leaders pay an additional cost cL = f(p(xp = xL), s(ΩxL,xlp:l∈A)),

∂cL∂p(xp=xL)

<

0, ∂cL∂s(•) > 0 that essentially forces them to internalize the damage to the party label from agent

signals that do not reinforce xL.

21

References

Adams, James and Samuel Merrill, III. 2008. “Candidate and party strategies in two-stage elections

beginning with a primary.” American Journal of Political Science 52(2):344–359.

Aldrich, John H. 1995. Why Parties? The Origin and Transformation of Political Parties in

America. American Politics and Political Economy Series. Chicago: University of Chicago Press.

Amorim Neto, Octavio and Fabiano Santos. 2001. “The Executive Connection: Presidentially

Defined Factions and Party Discipline in Brazil.” Party Politics 7(2):213–234.

Ansolabehere, Stephen and James M Snyder, Jr. 2000. “Valence Politics and Equilibrium in Spatial

Election Models.” Public Choice 103(3):327–336.

Belloni, Frank P. 1978. “Factionalism, the Party System, and Italian Politics.” In Faction Politics:

Political Parties and Factionalism in Comparative Perspective, Belloni, Frank P and Dennis C

Beller, eds., Studies in International and Comparative Politics, chap. 4, pp. 73–108. Oxford and

Santa Barbara, CA: Clio Press.

Bettcher, Kim Eric. 2005. “Factions of Interest in Japan and Italy.” Party Politics 11(3):339–358.

Carroll, Royce, Gary W. Cox, and Monica Pachon. 2006. “How Parties Create Electoral Democracy,

Chapter 2.” Legislative Studies Quarterly 31(2):153–174.

Carty, R Kenneth. 2004. “Parties as Franchise Systems: The Stratarchical Organizational Imper-

ative.” Party Politics 10(1):5–24.

Cox, Gary W. 1987. The Efficient Secret: The Cabinet and the Development of Political Parties

in Victorian England. Cambridge: Cambridge University Press.

Cox, Gary W. 2006. “The Organization of Democratic Legislatures.” In Oxford Handbook of

Political Economy, Weingast, Barry R. and Donald Wittman, eds., pp. 141–161. Oxford: Oxford

University Press.

Cox, Gary W. and Mathew D. McCubbins. 1993. Legislative Leviathan: Party Government in the

House. Berkeley: University of California Press.

22

Cox, Gary W, Frances M Rosenbluth, and Michael F Thies. 1999. “Electoral Reform and the Fate

of Factions: The Case of Japan’s Liberal Democratic Party.” British Journal of Political Science

29(1):33–56.

Cox, Gary W, Frances M Rosenbluth, and Michael F Thies. 2000. “Electoral Rules, Career Ambi-

tions, and Party Structure: Comparing Factions in Japan’s Upper and Lower House.” American

Journal of Political Science 44(1):115–122.

Cyr, Arthur. 1978. “Cleavages in British Politics.” In Faction Politics: Political Parties and

Factionalism in Comparative Perspective, Belloni, Frank P and Dennis C Beller, eds., Studies in

International and Comparative Politics, chap. 12, pp. 287–303. Oxford and Santa Barbara, CA:

Clio Press.

Dewan, Torun and Rafael Hortala-Vallve. 2011. “The Three As of Government Formation: Ap-

pointment, Allocation, and Assignment.” American Journal of Political Science 55(3):610–627.

Dewan, Torun and David P. Myatt. 2010. “The Declining Talent Pool of Government.” American

Journal of Political Science 54(2):267–286.

Di Palma, Giuseppe. 1977. Surviving Without Governing: The Italian Parties in Government.

Berkeley: University of California Press.

Duverger, Maurice. 1976. Les partis politiques. Paris: Librairie Armand Colin, 3rd ed.

Epstein, Leon D. 1967. Political Parties in Western Democracies. New York: Praeger.

Fukui, Haruhiro. 1978. “Japan: Factionalism in a Dominant-Party System.” In Faction Politics:

Political Parties and Factionalism in Comparative Perspective, Belloni, Frank P and Dennis C

Beller, eds., Studies in International and Comparative Politics, chap. 3, pp. 43–72. Oxford and

Santa Barbara, CA: Clio Press.

Giannetti, Daniela and Michael Laver. 2005. “Party Cohesion, Party Factions and Legislative Party

Discipline in Italy.” Paper presented at the Joint Workshop Sessions of the European Consortium

for Political Research, Granada, Spain.

23

Green-Pedersen, Christoffer and Kees van Kersbergen. 2002. “The Politics of the ’Third Way’:

The Transformation of Social Democracy in Denmark and the Netherlands.” Party Politics

8(5):507–524.

Groseclose, Tim. 2001. “A Model of Candidate Location when One Candidate Has A Valence

Advantage.” American Journal of Political Science 45(4):862–886.

Heller, William B. and Carol Mershon. 2008. “Dealing in Discipline: Party Switching and Legislative

Voting in the Italian Chamber of Deputies, 1988-2000.” American Journal of Political Science

52(4):910–925.

Heller, William B. and Carol Mershon, eds. 2009. Political Parties and Legislative Party Switching.

New York: Palgrave Macmillan.

Hu, Weiwei. 2013. “Career Ambition, Party Unity, and Party Policy Positioning: A Comparative

Examination of Intra-party Organization, 1945 - 2010.” Phd dissertation, Binghamton University

(SUNY).

Hu, Weiwei and William B. Heller. 2011. “Investing in Agreement: Party Organization, Leadership

Change, and Policy Positions.” Paper presented at the Annual Meeting of the Midwest Political

Science Association, Palmer House Hilton, Chicago, Ill., 31 March-3 April.

Huber, John D. 1996. Rationalizing Parliament: Legislative Institutions and Party Politics in

France. Cambridge and New York: Cambridge University Press.

Ignazi, Piero and Colette Ysmal. 1998. “Introduction.” In The Organization of Political Parties

in Southern Europe, Ignazi, Piero and Colette Ysmal, eds., pp. 1–5. Westport, CT, and London:

Praeger.

Indridason, Indridi H. and Christopher Kam. 2008. “Cabinet Reshuffles and Ministerial Drift.”

British Journal of Political Science 38(4):621–656.

Kam, Christopher, William T Bianco, Itai Sened, and Regina Smyth. 2010. “Ministerial Selec-

tion and Intraparty Organization in the Contemporary British Parliament.” American Political

Science Review 104(02):289–306. ISSN 0003-0554.

24

Kam, Christopher and Indridi H. Indridason. 2005. “The Timing of Cabinet Reshuffles in Five

Westminster Parliamentary Systems.” Legislative Studies Quarterly 30(3):327–363.

Laver, Michael. 1998. “Models of Government Formation.” Annual Review of Political Science

1:1–25.

Laver, Michael. 1999. “Divided Parties, Divided Government.” Legislative Studies Quarterly

224(1):5–29.

Laver, Michael. 2006. “Legislatures and Parliaments in Comparative Context.” In Oxford Hand-

book of Political Economy, Weingast, Barry R and Donald Wittman, eds., pp. 121–140. Oxford:

Oxford University Press.

Laver, Michael and Kenneth A. Shepsle. 1990. “Government Coalitions and Intraparty Politics.”

British Journal of Political Science 20(4):489–507.

Laver, Michael and Kenneth A. Shepsle. 1996. Making and Breaking Governments: Cabinets and

Legislatures in Parliamentary Democracies. Political Economy of Institutions and Decisions.

Cambridge and New York: Cambridge University Press.

Laver, Michael and Kenneth A. Shepsle. 1999. “How Political Parties Emerged from the Primeval

Slime: Party Cohesion, Party Discipline, and the Formation of Governments.” In Party Discipline

and Parliamentary Government, Bowler, Shaun, Farrell David M., and Richard S. Katz, eds.,

pp. 23–48. Columbus: Ohio State University Press.

Laver, Michael and Kenneth A. Shepsle. 2000. “Ministrables and Government Formation:

Munchkins, Players and Big Beasts of the Jungle.” Journal of Theoretical Politics 12(1):113–124.

McCubbins, Mathew D and Michael F Thies. 1997. “As a Matter of Factions: The Budgetary Impli-

cations of Shifting Factional Control in Japan’s LDP.” Legislative Studies Quarterly 22(3):293–

328.

McKelvey, Richard D. 1976. “Intransitivities in Multidimensional Voting Models and Some Impli-

cations for Agenda Control.” Journal of Economic Theory 12:472–482.

25

Mershon, Carol. 2001a. “Contending Models of Portfolio Allocation and Office Payoffs to Party

Factions: Italy, 1963-79.” American Journal of Political Science 45(2):277–293.

Mershon, Carol. 2001b. “Party Factions and Coalition Government: Portfolio Allocation in Italian

Christian Democracy.” Electoral Studies 20(4):555–580.

Mershon, Carol. 2002. The Costs of Coalition. Stanford, Calif.: Stanford University Press.

Michels, Robert. 1915. Political Parties: A Sociological Study of the Oligarchical Tendencies of

Modern Democracy. Hearst’s International Library.

Moreno, Erika. 2005. “Whither the Colombian Two-Party System? An Assessment of Political

Reforms and their Limits.” Electoral Studies 24(3):485–509.

Morgenstern, Scott. 2001. “Organized Factions and Disorganized Parties: Electoral Incentives in

Uruguay.” Party Politics 7(2):235–256.

Nicholson, Norman K. 1978. “Factionalism and Public Policy in India: The Vertical Dimension.” In

Faction Politics: Political Parties and Factionalism in Comparative Perspective, Belloni, Frank P

and Dennis C Beller, eds., Studies in International and Comparative Politics, chap. 7, pp. 161–

188. Oxford and Santa Barbara, CA: Clio Press.

Olson, Mancur. 1965. The Logic of Collective Action: Public Goods and the Theory of Groups.

Harvard Economic Studies. Cambridge, Mass.: Harvard University Press.

Osbourne, Martin J. and Al Slivinski. 1996. “A Model of Political Competition with Citizen-

Candidates.” The Quarterly Journal of Economics 111(1):65–96.

Panebianco, Angelo. 1988. Political Parties: Organization and Power. New York and Cambridge:

Cambridge University Press.

Plott, Charles R. 1967. “A Notion of Equilibrium and Its Possibility under Majority Rule.” The

American Economic Review 57(4):787–806.

Roback, Thomas H and Judson L James. 1978. “Party Factions in the United States.” In Faction

Politics: Political Parties and Factionalism in Comparative Perspective, Belloni, Frank P and

26

Dennis C Beller, eds., Studies in International and Comparative Politics, chap. 14, pp. 329–355.

Oxford and Santa Barbara, CA: Clio Press.

Saalfeld, Thomas. 2000. “Members of Parliament and Governments in Western Europe: Agency

Relations and Problems of Oversight.” European Journal of Political Research 37(3):353–376.

Sabatini, Christopher. 2003. “Decentralization and Political Parties.” Journal of Democracy

14(2):138–150.

Samuels, David J. and Matthew Soberg Shugart. 2010. Presidents, Parties, and Prime Ministers.

Cambridge: Cambridge University Press.

Schofield, Norman. 1983. “Generic Instability of Majority Rule.” Review of Economic Studies

50(4):695–705.

Schofield, Norman. 2007. “The mean voter theorem: necessary and sufficient conditions for con-

vergent equilibrium.” Review of Economic Studies 74(3):965–980.

Schofield, Norman and Itai Sened. 2005. “Multiparty Competition in Israel, 198896.” British

Journal of Political Science 35(04):635. ISSN 0007-1234.

Schofield, Norman and Itai Sened. 2006. Multiparty Democracy: Parties, Elections and Legislative

Politics in Multiparty Systems. Cambridge and New York: Cambridge University Press.

Shepsle, Kenneth A. 1979. “Institutional Arrangements and Equilibrium in Multidimensional Vot-

ing Models.” American Journal of Political Science 23(1):27–59.

Strøm, Kaare. 1990. Minority Government and Majority Rule. Cambridge: Cambridge University

press.

Strøm, Kaare. 2004. “Parliamentary Democracy and Delegation.” In Delegation and Accountability

in Parliamentary Democracies, Strøm, Kaare, Wolfgang C. Muller, and Torbjorn Bergman, eds.,

pp. 55–106. Oxford: Oxford University Press.

VanDusky-Allen, Julie and William B. Heller. 2014. “Bicameralism and the Logic of Party Orga-

nization.” Comparative Political Studies 47(5):715–742.

27

Zuckerman, Alan S. 1979. The Politics of Faction: Christian Democratic Rule in Italy. New Haven:

Yale University Press.

28