A Theory of Political Party Organization

A THEORY OF PARTIES: 1 POLICY DEMANDERS, LONG COALITIONS AND THE ELECTORAL BLIND SPOT Kathleen Bawn, UCLA Marty Cohen, UCLA David Karol, UC Berkeley Seth Masket, University of Denver Hans Noel, UCLA and CSDP, Princeton University John Zaller, UCLA May 14, 2006 This paper offers a theory of political parties that explains how their various branches -in the government, in the electorate, and in organization -- work together to structure political conflict. Its basic posture is that parties are a conspiracy of intense minorities against the less attentive and involved. Yet, as a by-product of their wish to govern, parties offer a degree of responsiveness and accountability to all voters. The extent of parties’ responsiveness and accountability depends on how closely voters monitor them. Parties wish to cede only as much to public opinion as necessary to win control of government via fair elections, so they do what they can, via agenda control, secrecy, and other devices, to make monitoring difficult. Nominations for office are the fundamental act of parties. Choice of nominee reflects the balance of forces within a party, structures electoral politics, and sets the agenda of government. The fact that parties sometimes struggle in primary elections to control nominations does not make nominations less important. When party interest groups control candidate choice, and especially when they enjoy the power of de-nomination, they insist on candidates who deliver policy benefits to coalition members. When candidates gain control of their own nominations, as sometimes occurs, delivery of partisan benefits slows and parties themselves may wither. Nothing in our theory challenges the important argument, widely accepted in the literature, that politicians build party institutions, both inside and outside of government, to advance their goals. But the goals of politicians are, in our theory, mainly the goals of the groups and activists that nominate them. When new groups enter the party coalition, or when old groups make new demands, the goals of office-holding politicians change. Office-holders, most notably presidents, have some capacity to shape party coalitions, but our theory holds that groups and activists are generally dominant. Thus, in contrast to recent theorizing, which tends to view party institutions as creatures of partisan officeholders, we view partisan office-holders as the creatures of their group and activist backers. Even for the current period, we propose a group- and activist-centered theory of parties rather than a candidate-centered one. Please do not cite without obtaining the most recent draft. We thank Thomas Schwartz for saving us from many but not all of our errors, and Casey Dominquez, for incisive comments on a related paper. 1 1 Although we emphasize that parties are anything but the humble servants of democracy, our theory is consistent with the widespread view that, as put by E. E. Schattschneider, “modern democracy is unthinkable save in terms of parties (p.1).” In our theory, parties serve voters for the same reason that profit-hungry corporations serve consumers -because they need public support in order to achieve their own goals. Under the right conditions, parties provide a high level of service to voters. Whether, over the long haul, the level is high enough is a difficult normative problem that we do not engage in this paper. Our purpose here is to explain party behavior. And since, as we hold, the aim of party interest groups and activists is to maximize their share of policy benefits, our explanation focuses on that motivation. The building blocks of our theory are, as the reader will quickly discover, familiar. It builds above all on Schattschneider’s definition of a party as “an organized attempt to seize control of the government,” making this notion applicable to more recent theorizing and to current party practice. The theory attempts, inter alia, to reconcile the competing arguments of David Mayhew’s Electoral Connection, which explains why office-holders have little use for parties, and Thomas Schwartz’s and John Aldrich’s “Why Parties?” arguments, which explain why they do. It seeks also to explain why parties sometimes nominate relatively extremist candidates and why they sometimes offer candidates closer to the median, as documented especially by Ansolabehere, Stewart and Snyder in their paper on “Candidate Positions in Congressional Elections.” David Rhode's theory of Conditional Party Government sits comfortably within our argument, as does recent research on the importance of Rules Committee agenda-setting in legislative politics. The theory explains why the current system of presidential nominations, which seems to some to be over-run by special interests, is actually an example of a strong party process, as maintained by Cohen, Karol, Noel, and Zaller in their paper on “Beating Reform: The Resurgence of Parties in Presidential Nominations.” The paper is a preliminary endeavor. Although organized in a deductive manner, its conclusions are not formally proved. It refers only to the United States. Although we make use of framework that Bawn and Rosenbluth (2003, 2006) developed to study the multi-party systems of Europe, we make no effort to explore cross-national comparisons here. We nonetheless believe that our theory has value for pointing out connections among widely dispersed party activities and creating a unified framework for explaining them. The next section of the paper develops our theory of parties, first as a story, then with somewhat more rigor. The middle section of the paper presents empirical evidence relevant to the theory. The evidence is not meant to be a conclusive test; its consists partly of sketches and stylized facts. Our aim in this brief paper is lay out the theory as clearly as possible and to establish its empirical plausibility. In conclusion we briefly evaluate the normative implications of the kind of party we have theorized. A THEORY OF POLITICAL PARTIES Our argument builds on the “Why Parties?” argument of Thomas Schwartz and John Aldrich, which we take to be the leading theoretical position on why parties form. Their parties emerge from a mythical legislature that initially has no ideological structure. 2 Suppose, following Schwartz, that the legislature has three members and an agenda that consists of three bills having the following payoff structure: Bill A Bill B Bill C Mr. 1 4 4 -9 Mr. 2 4 -9 4 Mr. 3 -9 4 4 Schwartz notes that if the bills are considered one at a time, all will pass, because each can individually command majority support. However, this result is not optimal, because each legislator will win twice (for a total gain of 4 + 4 = 8) and lose once (for a loss of 9), leaving each person worse off than defeating all three bills. But suppose that Mr. 1 and Mr. 2 agree to jointly support Bill A and to oppose the other two bills. They will each make a gain of 4 on that bill while Mr. 3 will be stuck with a loss of 9. Although this is scarcely fair, the temptation for politicians to form such a majority coalition will be, as Schwartz argues, great. And any two legislators, having cooperated once, will be tempted to keep cooperating, thereby forming a “long coalition” over many mutually advantageous issues. Schwartz generalizes his argument to a larger body of legislators, where the temptation to form a majority coalition that wins on all issues of mutual advantage to coalition members is again expected to be irresistible. The losers, as Schwartz continues, take their grievances to the voters in an effort to win more seats and take over as the new majority. Finally, voters respond to the two coalitions as appropriate to their interests, thereby generating a full-fledged party system. Schwartz takes this set of ideas as an explanation for why parties exist, and Aldrich embeds it in a larger argument in his book length treatment of Why Parties? Aldrich notes in particular the value of parties for mobilizing voters and stabilizing the careers of politicians. He states his argument as follows: My basic argument is that the major political party is the creature of the politicians, the ambitious office seeker and the officeholder. They have created and maintained, used or abused, reformed or ignored the political party when doing so has further their goals and ambitions. The political party is thus an “endogenous” institution – an institution shaped by these political actors. (p. 4) The goals of politicians are those of Richard Fenno’s famous study of Congressmen in Committees, namely, to make good public policy, to enjoy a long and stable careers, and to attain power and prestige in government. As Aldrich fleshes out his theory for the contemporary period, he adds policy activists who pressure party office-holders to take ideologically extreme positions. Ambitious politicians, who need to win both a primary election dominated by activists and a general election dominated by centrist voters, must then position themselves as best they can 3 between the two groups. Office-holders build institutions in Congress that permit them to achieve their goals of partisan legislation, and they build party institutions that aid in reelection. Party institutions in the current period are, thus, “in service” to party officeholders and candidates. Aldrich’s theory, with the ambitious politician at its center, does a fine job of organizing salient facts about the party system. It is a landmark in the study of political parties. We believe, however, that a more satisfying account of party politics can be achieved by shifting the focus of the theory from the politician to the groups and activists that ally themselves into the party coalitions. Our theory deals with many of the same facts, and often through a similar theoretical lens, but ends up giving a different impression of the nature of parties. Long Coalitions Outside the Legislature: A Story Our theory is also best introduced by imagining a state of nature in which no parties yet exist. In our founding myth, society does not have a government, but it is about to elect its first president via a plurality election. The president is a dictator, in the sense that he makes policy decisions unilaterally (there is no legislature in this myth). But he serves for a fixed short term, with re-election possible but not guaranteed. The society has a few thousand voters and is mostly contented, but it contains five “intense minorities.” The intense minorities are groups of persons eager for policies that disproportionately benefit them but whose costs are borne jointly by the whole society. For example, an intense minority may be shepherds who want a high tariff on wool imports. The tariff increases the prices shepherds can charge for their home-grown wool, which are borne by all members of society. Despite the society’s modest size, most voters know little about the intense minorities and even less about their arcane policy goals. As the election approaches, one of the minority groups – let us say, the shepherds – proposes to work together to elect one of their own, who would naturally be favorable to a high tariff on wool imports. The shepherds are not a rich or numerous group, but in an otherwise unstructured electoral environment, with most voters uninterested in the election, their chance of winning seems high. But then another intense minority, the teachers, notices the shepherds. The teachers calculate that they could easily outspend the shepherds to elect a teacher, whose top priority would be new school construction. At this point, the other three minorities recognize that, if they banded together, they could outspend the teachers and the shepherds and thereby control the office of dictator on behalf of their intense concerns. As soon as the three groups begin to work together, however, their coalition encounters problems. The coffee growers want a new four-lane highway to their remote region in order to increase market access, but the other two groups, the clergymen and auto workers, are dubious. For one thing, it would require a tax increase that the voters, though generally inattentive to politics, would likely notice. For another, the groups 4 themselves would also be burdened with the tax. Several such concerns arise, but the three groups bargain them out. For example, everyone agrees that a two-lane highway should suffice for the coffee growers. The clergy’s plans to promote religious orthodoxy and the auto workers’ framework for collective bargaining rights are similarly scaled back from what each group had originally envisioned. Worried that voters may be suspicious of a candidate sponsored by special interests, the groups give their coalition a catchy name – the Justice Party – and try to keep generally in the background. But to insure support from the least engaged members, the groups nominate a church-going coffee drinker who is married to an auto-worker. When the teachers and shepherds realize that they face the united opposition of the other three intense minorities, they form their own coalition, which they call the Fairness Party. They bargain out their differences and also work hard to find an appealing candidate. Succeeding in this, the Fairness Party wins the election. Upon taking office, the president immediately raises the wool tariff and builds a large new school house, including a lavish faculty lounge. Everything up to this point in our creation myth would likely occur if the country were electing a dictator for life. However, the existence of regular elections creates an incentive for on-going politicking. So, before the paint in the new teacher’s lounge is dry, the losing candidate of the Justice Party attacks the president for favoring special interests and continues his attacks into the next election campaign, in which he easily defeats the incumbent. The second president is more savvy than his defeated predecessor. He pays little attention to the interests that sponsored him and spends most of the national budget on a project of interest to all voters -- fireworks for the Independence Day celebration. This enrages the three intense minorities of the Justice Party, who withdraw their support at the next election, but the incumbent continues to claim the mantle of the Justice Party and wins re-election anyway. Learning a lesson, the Justice Party takes much more care in future years to select nominees with proven loyalty to party goals. Our founding myth explains why parties are likely to form in any unregulated electoral setting and how they might be expected to work. The main reason for party formation is that groups with intense policy concerns– which could derive from idealism as well as self-interest – see advantage in organizing to control government by controlling the nomination and election of office-holders. We hope that readers will agree that our founding myth has both internal coherence and many realistic elements. The most important of these elements – the existence of intense minorities and a non-intense majority – is a persistent and highly significant feature of political life. Our founding myth shares the basic “long coalition” logic of Aldrich and Schwartz’s founding myth, yet differs in three significant ways. First, parties form outside the legislature, and this happens before any election has occurred. Second, in the process of forming a long coalition, the groups bargain and compromise with each other. Finally, the mostly inattentive and uninvolved majority of voters critically affects the 5 success of the party as long coalition. Parties try to strategically anticipate the actions of inattentive majority, leading them to further compromise their founding goals. How, and how much, to compromise becomes a source of conflict with the long coalition, in particular between the group at the party’s core and the candidates who compete under its banner. Further discussion of long coalitions and compromise requires restating parts of the argument in a somewhat more formal and abstract way. Parties, Candidates, Voters and Compromise: A Model We begin our model with the following fundamental assumption: Citizens become active in politics in proportion to their expected gains from activity. We allow that the gains from political action may be partly subjective. Two citizens may have the same policy choice -- e.g., freedom of choice on abortion -- and yet vary greatly in the gains they experience from a government policy that is pro-choice on abortion. Thus, citizens who become active in politics will be people whose preferences are far from the policy status quo and who care intensely about those policies. When the policies of a given society are far from what a majority of citizens strongly wants, the majority will become active and, being a majority, probably achieve its goals. Those who remain active in politics over a long period are likely to be minorities who cannot so easily achieve their deeply felt preferences. The more extreme an individual’s preferences in relation to what the majority wants, the greater the difficulty in achieving them and therefore the greater the need for long-term concerted action. We call the individuals who engage in such action intense minorities. The goals sought by intense minorities may be anything that can be achieved through government action and that a majority might for any reason oppose. These goals include: 1. Well-paying, low-effort, and dependable jobs on the government payroll for persons having low skills. 2. Policies that give special advantage to a particular business, such as tariffs, price supports, tax incentives, aggressive regulation of competitors, or protection from regulation that a majority would favor. 3. Redistribution of wealth in the interest of a particular notion of fairness or, alternatively, protection against redistributive policies that a majority would favor. 4. Policies that promote a particular vision of morality, such as sexual abstinence outside of marriage or homosexual marriage. 5. Policies that promote a broad ideological vision of society, such as socialism, slavery, theocracy, free market capitalism, liberalism, or conservatism. 6 Depending on policy details, one can readily imagine that a majority of citizens -possibly an intense majority of citizens -- would oppose any of these policies. Yet some individuals may, for reasons of either material self-interest or high-minded morality, intensely favor them. Such intense minorities are at the center of our theorizing. At some points, we shall refer to these intense minorities as policy demanders. Policy demanders may, of course, be majorities, but for reasons already given, majorities do not usually need to engage in sustained political action on behalf of their demands. Hence, in our theory policy demanders will almost always be intense minorities. As in the Aldrich-Schwartz example, we find it useful to talk about the benefits and costs of policies in terms of numerical values. This makes obvious sense when discussing monetary benefits, such as government services or the taxes that pay for them, but we shall also discuss policies that involve only moral values, such as the value of a pro-life or pro-choice abortion policy, in terms of numerical values. We do this because we wish to have a common metric for measuring the very different kinds of policies that different intense minorities favor or oppose. This metric is the personal utility created by the policy for the citizens who may like or dislike it. In the Aldrich and Schwartz examples, the costs and benefits of projects were exogenously fixed -- a legislator either gained 4 units or lost 9. The fixed size of the projects limits our ability to discuss compromise, which is an essential feature of party politics. If we want to talk about how groups compromise to form parties and how parties compromise to win elections, it is helpful to allow the size of the project to change. In the founding myth, compromise occurs when the groups scale back their demands: from a four-lane to a two-lane highway, a high to a medium wool tariff, a luxurious to a comfortable teachers’ lounge. Bawn and Rosenbluth (2003, 2006) incorporated compromise into the Aldrich-Schwartz model that focuses on compromise between groups. We will summarize their framework here in the context of our founding myth, and then extend it to incorporate voters. Continuing the five-group example from our founding myth, let xi denote the scale of the i-th project for i =1...5. That is, let x1 be the wool tariff, x2 is the school construction budget, x3 the coffee growers’ highway, x4 is religious orthodoxy, and x5 protection of the right to bargain collectively . The benefits of each project i, B(xi), accrue only to group i, while the costs C(xi) are shared evenly among all members of society, so that each groups share is 1/n of the total. For convenience, assume that 2 benefits are linear, that is, B( xi ) = xi , and that costs are quadratic, or C ( xi ) = xi , for all projects. From the point of view of the shepherds, the ideal scale of the highway is that which 1 maximizes their benefits net of their share of costs, that is B ( x1 ) − C ( x1 ) . It is easy to n show that, given this formulation, the shepherd’s optimal level of benefit, at which their n marginal benefit equals their share of marginal cost occurs when x1 = . The shepherds 2 therefore ideally prefer to spend n/2 on the highway. However, what is ideal for 7 shepherds is not ideal for their coalition partners, the teachers. The teachers would ideally prefer to spend nothing on highways, but would like to spend n/2 on school construction. In the process of forming the Fairness Party, the teachers convince the shepherds to settle for a somewhat lower tariff, one that takes into account (in economist’s jargon, “internalizes”) the teachers’ share of costs of the tariff. This would n 2 reduce the tariff to the level that maximizes B ( x1 ) − C ( x1 ) , or x1 = . Similarly, the n 4 shepherds prevail upon the teachers to accept a school construction budget that is larger than socially optimal (larger than if all costs were considered) but smaller than what the teachers regard as ideal. Bawn and Rosenbluth argue that parties as long coalitions generally promote this kind of compromise among the groups they contain. Furthermore, they show (BR 2003) that the kind of competition between two long coalitions (like our Justice Party and Fairness Party) can be sustained as a Nash equilibrium under plurality electoral rules. 2 Bawn and Rosenbluth’s main concern was with compromise among the intense minorities that can form long coalitions, and how the terms of compromise vary under different party systems. Our interest here lies more in other types of compromise, specifically the kinds of compromises that long coalitions may make with unattached voters and with individual candidates. A spatial model is useful for discussing compromise, as it draws our attention to the distances between actors’ ideal policies and other policies that they might accept. Strictly speaking, the pork barrel model sketched so far is a spatial model, with as many dimensions as there are groups, and with ideal points organized in a particular way (each group’s ideal point is 0 on every dimension but its own, where it is n/2.) The high dimensionality of the pork barrel model is important to our overall argument that long coalitions are socially constructed, not simply a natural clustering of the truly likeminded (e.g., as in Krehbiel 1993.) Even though we assume that the true space of policy choices has high dimension, it is reasonable to expect that observed political competition will occur in a space of much lower dimension. The reduction of observed dimensionality will be a direct by-product of the politics of long coalitions. Specifically, as the Fairness and Justice Parties continue competing over a series of elections, campaigns will regularly feature wool tariffs and education spending linked together. A pundit or a philosopher who argues that both high tariffs and well-appointed teachers’ lounges are directly implied by fundamental normative principles is likely to be cited often by Fairness partisans, more often than other pundits and philosophers whose work is not so relevant to the problems of political practioners. 3 The same of course will be true for writers who manage to link the pet projects of coffee growers, the clergy and 2 Bawn and Rosenbluth 2003 also showed, not surprisingly, that competition structured around just two long coalitions is not in equilibrium under proportional representation, although other systems with smaller long coalitions are. 3 Noel 2006 argues that this may even precede the formation of the coalition, and so help determine which groups align. 8 auto workers into a coherent ideological framework. 4 People will begin to believe that preferences for high tariffs and those for fancy teachers’ lounges naturally go together. 5 Thus, even though the true political space has (at least) five dimensions, party and electoral discourse will be dominated by two dimensions, or perhaps only one. As illustrated in Figure 1, there will be a Fairness dimension which will be generally understood to imply high tariffs and educational infrastructure, and a Justice dimension, implying transportation, religious values and the right to bargain collectively. An easy concrete way to think of the each dimension is that it captures some fixed ratio of benefits to each of the groups in the ideological coalition. The fixed ratio is hammered out when the long coalition is formed (e.g. in the initial bargaining between shepherds and teachers over the party agenda). Assume that the ratio is unity, so that each intense minority that participates in the long coalition benefits equally. (Relaxing this assumption would have no impact on the main argument.) Moving up the y axis, for example, implies increasing both the tariff and education benefits in roughly equal proportions. 6 Under the assumption that the party groups are intense minorities, their gains in “Fairness” represent costs to the population as a whole. The positions of the five groups/two long coalitions are indicated in Figure 1. The ideal point of the shepherds and teachers, S-T occurs at the point where the marginal benefits of the wool tariff and teachers’ lounge equal the two groups’ joint share of marginal cost, that is each group gets its project at a scale of n/4. Similarly, the ideal point of the coffee growers, clergy and autoworkers implies these groups’ (jointly) optimal levels of their pet projects. As in Bawn (1999), each group will sometime take positions on individual issues that look ideological, in the sense of not being consistent with the groups’ material interest (as for example when the teachers support the wool tariff.) The important point is that, given the existing long coalitions, the group’s ideal point is the one which maximizes its net benefits. There are several reasons to think that long coalitions benefit from ideological discourse that binds them together. One is that long coalitions suffer from uncertainty. As unforeseen events occur (the cotton gin is invented, changing dramatically the international market for wool) how should the coalition react? The shepherds might argue that they should continue to enjoy the same level of well-being as before and that the Fairness Party’s platform should adjust to accomplish this. The teachers might take a different perspective. But both groups are hurt if their coalition crumbles. Ideological arguments that give the groups a common understanding of how to react to foreseen circumstances can keep the coalition from falling apart in the same way that corporate culture supports cooperation within firms under incomplete contracts (Kreps 1990). 5 This reduction of dimensionality could take place in the legislature, in political campaigns, or in general elite or cultural discourse outside the party system. If it were to occur in the legislature, the likelihood of reduction to a single dimension would be higher when the legislative agenda is controlled by a long coalition (Cox and McCubbins), which would structure legislation so that roll call votes fall on along a single dimension. Wright and Schaffner (2003) and Masket (2004) demonstrate that roll call votes show higher dimensionality in state legislatures when parties play less of a role. The demonstration that agenda control reduces dimensionality in legislatures was first made by Sndyer (1992). However, Noel 2006 shows that unidimensionality appears in the extra-legislative domain of pundit and intellectual discourser as well and may be the source of the dimensional structure that appears in the party system. 6 Hinich and Munger (1996) argue that ideologies constrain the set of policy choices in precisely this way. 4 9 Figure 1 Fairness S-T a a′ V b’ b C-C-A Justice 10 What of the unattached voters, those not associated with any intense minority? They pay the costs of wool tariffs, teachers’ lounges and highways to coffee-growing areas, but experience no benefits. In reality, of course, some of the projects will probably provide positive externalities (happy teachers performing better in the class room, use of the highway). Externalities notwithstanding, unattached voters will prefer much lower levels of all projects. The important point here is that the ideal point of the unattached in the space of political discourse is close to the origin. The notion that we are trying to capture is that the unattached voters are uninformed (in way that will be made precise just below) but they are not systematically deluded. Just as with the groups, the voters’ preferences may derive from material interest or from something else. That is, the unattached voters’ position near the origin reflects not only tax burden, but the nonmonetary costs of, for example, laws requiring observance of various religious rituals, holiday and taboos. If the unattached voters know what each party stands for, they will vote for the party with the lowest “cost” platform (with costs broadly construed), creating pressure for both parties to scale back the projects they were originally created to promote. But a big part of the reason these voters are unattached is that they generally pay little attention to politics. If the unattached pay little attention, rarely distinguish between the two parties’ policies, no matter how outrageous or far from their own actual if non-intense preferences, then they will have little impact on the parties’ platforms. This level of unconditional ignorance and disaffection is too extreme. We do not contend that the unattached voters never notice what the parties stand for or that their views are entirely independent of the ideological structure. A more reasonable assumption is that unattached voters notice, and react to, big differences in policies, but not small ones. The inability to perceive small differences is usually depicted in spatial models as “thick” indifference curves. Consider the two circles in Figure 1, both centered at the origin that, is at the unattached voter’s ideal point. If voters are sufficiently informed to perfectly discriminate between all policy alternatives, each circle would constitute an indifference curve (being the set of all points equally far from the ideal point.) But if the low information of unattached voters means they are unable to perceive small (or even medium) differences, then their indifference sets will be bands, rather than curves. Figure 1 depicts the first two indifference bands of the unattached voter. That is, she is unable to discriminate between any of the policies in the interior of the smallest circle (including her ideal point.) In the eyes of the unattached voter, these are all reasonable policies; she would not quibble with any of them. The next indifference set consists of all the points between the two circles. The voter recognizes that both policies a and b in this region are more extreme than she would like, but is unable to recognize that Point a (too much Fairness) is actually closer to her ideal point than Point b (too much Justice.) 11 If low information creates these wide bands of indifference for unattached voters, each party has an incentive to scale back benefits for its core groups so that its policy proposal lies just on the edge of the unattached voters’ highest indifference plateau – that is, to a point like a′ or b′ lying just on the edge of the innermost circle. A party that scales back less can expect to be defeated. That is, if Fairness scales back its wool tariff and school building proposals to point a′, but Justice only scales back its highway, religion and collective bargaining proposals to b (or, worse, insists on its ideologically correct ideal point of CCA), Justice will lose. Their is no need for either party, however, to scale back beyond the point where unattached voters will notice any difference. For this reason, we can think of the unattached voters’ highest indifference set as “the electoral blind spot.” As long as parties stay within the electoral blind spot, they are effectively free to set whatever policy they want. They have nothing to gain from further compromise, nothing to lose from sticking to their guns. The size and shape of the electoral blind spot varies from one election to the other. Sometimes voters are more attentive than others, and they are more attentive to some issues than others. This variability means that parties can never be certain about the precise location of the blind spot’s edge. Much of the political activity we observe before elections (polling, focus groups, tweaking of policy positions, trial balloons on various issues) seems motivated by the desire to see how much an ideologically extreme party or candidate needs to compromise with uninformed centrists, that is, by the desire to figure out where the edge of the electoral blind spot is. Uncertainty about the size of electoral blind spot has the important consequence of creating conflict between the long coalition and the individual candidate who wants the job of representing it. Assume that the Fairness Party’s candidate is a shepherd named Joe. Like the other shepherds, Joe sincerely wants the party to defend its core principles, such as protecting consumers from the adverse health effects of ugly, itchy imported wool. He wants Fairness to win because he truly believes in principles of Fairness, but, unlike the other shepherds, and other groups in the Fairness party, Joe also wants to win office for its own sake. Both Candidate Joe and the Fairness Party are thus willing to compromise (scale back the policy platform) in order to win the election, and if they knew exactly how much compromise was needed, they would not be in conflict. When there is uncertainty, however, the optimal amount of compromise reflects a trade-off between increasing the probability of winning (by moderating) and increasing the pay-off from winning (by sticking to principles.) The candidate and the party will systematically differ about how to optimally balance these competing goals. Specifically, let x be the platform of the Fairness Party, and y the platform of Justice. As above, we can think of x as measuring the size of the tariff (and also the teacher’s 12 lounge budget.) 7 Let U be the utility that the Fairness Party (more precisely, any given intense minority in the Fairness long coalition) receives from the policy implemented after the election. Let q(x,y) be the probability the Fairness Party wins the election. We are interested in the range of x where the party’s uncertainty about the size of the electoral blindspot is relevant, that is, where q is strictly between 0 and 1, where higher levels of x are better for the intense minorities ( U '( x) > 0 ) but imply a lower probability of winning ( q '( x) < 0 .) It will also be the case that U ''( x ) < 0 and that q ''( x ) < 0 . The negative second derivative of utility is implied by standard benefit and cost assumptions, and is necessary for ideal points to be well-defined. the positive second derivative of q follows from the fact that, as a probability, q is bounded below by 0. The party’s optimal choice of x solves max q ( x)U ( x) + (1 − q ( x ))U ( y ) . 8 x The first-order condition for this problem tells us party’s optimal policy occurs when U '( x) U ( x) − U ( y ) = q( x) (1) q '( x) In contrast, Joe’s optimal choice of x takes account of the value he places on winning office, V >0 max q ( x)(U ( x ) + V ) + (1 − q ( x ))U ( y ) x and implies a different first-order condition U '( x) q( x) (2) q '( x) Comparing equations 1 and 2, and bearing in mind that q '( x) < 0 , U ''( x ) < 0 and q ''( x ) < 0 , it is clear that the party’s ideal policy will be higher than Joes. Or, put the other way, that Joe will want to compromise more than the party does. U ( x) − U ( y ) + V = So far, this part of our story has pitted the individual candidate Joe against the party from which he emerged. In reality, Joe would have some allies within the party, others who value winning per se, and who would want to moderate more than the core party coalition. Joe’s allies would generally include his entourage, consultants who managed his campaign and wish to continue earning a living this way, and anyone who expects to receive patronage. This theorizing highlights what, as we believe, political parties try to do. They try to nominate candidates and enact policies that are as far from the ideological center of the political space as voters will easily accept. They do this in order to maintain the long 7 Concretely then, U ( x) = x1 − 1 5 2 ∑ xi where x = x1 = x2., It is worth emphasizing that our basic n i =1 theoretical framework remains constant throughout the various parts of our argument, but the extra level of detail about the incidence and shape of costs and benefits is not needed for the present point. 8 Because we are looking at the problem from a decision theoretic point of view, we supress q’s dependence on the other party’s platform, y. 13 coalition and keep the intense minorities mobilized. Uncertainty about what kinds of policy choices will and will not rouse the sleeping giant of the uninvolved majority complicates this task and creates systematic conflict between the party and its incumbent office holders. In our model, the essential party activity is bargaining and deliberation among members to find candidates and measures that can unite the party and yet attract enough voters to win in the fall election. This corresponds to key claims of existing empirical literature. For example, Polsby and Wildavsky write of the pre-reform presidential nominating system, “Unless party leaders achieve a consensus among themselves, chances are diminished that they will be able to elect a President (p. 71-72).” Most often, as they add, this is a candidate who is no one’s first choice but many people’s second choice. From a different theoretical perspective than ours, Cox and McCubbins (2005) maintain that it is the task of party leadership in the House to unite members of the party caucus around a position that is away from the median of the House -- which may be presumed to be near the national center of opinion -- and yet to be attractive to voters with the “brand label” thereby created. It is not impossible in our model that a party will give voters exactly what they want. This is especially likely to occur immediately after a new party takes office, as it happily concedes to voters popular policies that the outgoing party had blocked and that its own intense groups may not oppose. However, our claim is that parties do not generally aspire to be tribunes of the people. Indeed, even when voters want something that is agreeable to the governing party, the party may slyly give voters a bit more of it than they really want because this is what the party wants. Because parties routinely aspire to non-majoritarian policies, one expects them, as noted earlier, to be secretive about their internal processes and exactly what actions they are taking. This secrecy is most overtly exemplified by the image of the “smoke-filled room” at party nominating conventions and the artful legislative rules of the House Rules Committee. By such means, parties attempt to enlarge the electoral blind spot. But active efforts at hiding are often not necessary given the low level of interest on the part of most voters. Our view of voters requires further elaboration. We have already said that we follow Aldrich in the view that citizens become involved in politics in proportion to expected losses in political utility from government policy. Given that, in practice, most American voters are centrist or non-ideological, it follows that the bulk of voters develop only limited capacity to discern ideological differences among candidates. This view is controversial. As V.O. Key famously argued against those who denigrate the capacity of the citizenry, “voters are not fools.” Many have taken up Key’s contention, arguing that voters may rely upon heuristics, cues from trusted group leaders, values, or just common sense to reach good decisions (for reviews, see xxxx). 14 Our view is certainly not that voters are fools. To the extent that they pay attend to politics, or that circumstance brings relevant information to their attention, voters can compel the parties to be responsive to their preferences. But to the extent to which they remain uninformed, parties will attempt to take advantage of them and often succeed. No matter how generally wise voters may be, they cannot consistently outfox the sophisticated efforts of parties to exploit their ignorance. Or so our theory critically assumes. Four recent studies point up rather clearly the potential for parties to exploit voter lack of attention and information: • In a recent study, psychologists asked naive subjects to judge which of two faces suggested greater competence. The faces were those of two candidates in Senate elections. It turned out the relative differences in perceived confidence among subjects in the study correlated at .55 with the actual vote difference between the two candidates and correctly predicted 70 percent of the races. (Toidorov, Mandisodza, Goren, and Hall, 2005). These results leave plenty of room for issues to matter, but also plenty of room for voters to be exploited by parties that choose their candidates carefully. • Imagine a political world in which the performance of the economy in the last six months before the election heavily determined the outcome of presidential elections and in which economic performance was mostly random. It should be clear that parties in this situation would have only a limited incentive to give voters what they want, since their electoral fate would be largely determined by the luck of the economic draw just ahead of the election. Insofar as parties would exert themselves at all, it would be only at the end of their terms. As Achen and Bartels (2005) show, these conditions are not terribly far from reality. Late economic performance explains about half the variance in election outcomes and, to the extent parties make systematic efforts to control it, they do so mainly at the end of their terms. Parties in the Achen-Bartels world of “musical chairs” still have incentives to respond to voter preference, but the more voters respond to late and random economic performance, the weaker the incentive of parties to do so. • In analyses of the 2000 election, two studies find that small differences in relative volume of advertising have strikingly large effects. Johnston, Hagen, and Jamieson (2004) find that Bush came from behind and won the election on the strength of a roughly 6:4 spending advantage in battleground states in the last two weeks of the campaign. An earlier lead by Gore, built on a roughly similar advantage in advertising volume, was wiped out by Bush’s better-timed blitz. In a more technically sophisticated analysis of the same election, Huber and Arceneaux (2006) find even larger effects. As they write: The amount of advertising and the costs of the advertising necessary to change the electoral outcomes in these states are astonishingly low. This is perhaps not surprising for Florida, New Mexico, Wisconsin, Iowa, and Oregon, states that were decided by less that .25% of the 2-party vote. But the potential importance of relatively minor shifts in advertising does not stop there. Arkansas or Maine 15 could have been swung, to Gore and Bush respectively, for less than a million dollars. Considering that the Bush campaign spent nearly $11 million dollars relatively ineffectually in California alone, it is significant to recognize that simply having reallocated that money to New Mexico, Wisconsin, Iowa, Oregon, Minnesota, Maine, Michigan, and Washington could have turned 8 of the 9 states Gore won on this list into ties for about $9.3 million. On the other side, if the Bush campaign’s allocation had remained constant, for about $9.7 million Gore could have won Florida, New Hampshire, Missouri, Ohio, Nevada, Tennessee, and Arkansas by simply reallocating some of his $13 million “surplus” in Pennsylvania, Michigan, and Washington. • As Zaller (2004) shows, the voters most responsive to the effects of economic performance, and also the voters most likely to shift from one election to the next, are low information voters. A variety of evidence thus suggests that voter inattention to politics creates an opportunity for clever parties to exploit. The extent to which they are actually able to do so, as we argue below, is substantial but not infinite. And it requires substantial skill as well to create the united front behind the right candidates and measures. One must not press the notion of anti-majoritarian parties too hard. Parties must win elections and voters are not fools. Parties must therefore deliver some real benefits to popular majorities in order to gain the government control they covet. In many cases, the policies parties wish to deliver may be policies that a majority wants. Examples are the Social Security program passed by Democrats in the 1930s and the Republican welfare reform of 1996. Parties must also take responsibility for the performance of government while they are in office. In these important ways, parties foster popular control of government. Our point, however, is that parties do not exist to serve as instruments of popular control of government. They exist to work for the policies that their members, for whatever reasons, most prefer. To understand their behavior, one must focus on what actually motivates them, not what occurs as a by-product of their exertions. Our parties are quite different creatures than those of Schwartz-Aldrich. Where Aldrich views party as institutions that are the endogenous creation of ambitious politicians, we view ambitious politicians, including their policy positions and the institutions they create to pursue them, as the endogenous creation of the larger party coalition. Aldrich emphasizes that party institutions were “in service” to office-holders, helping them to achieve their goals. Our parties, in contrast, strive mightily to control their office-holders — by selecting them to be loyal to party goals, by making them dependent on its electoral service networks and threatening apostates with de-nomination. By these means, the party coalition normally but not always succeeds in controlling its politicians. Other more specific points of similarity and contrast are: • The most fundamental similarity between our model and Schwartz-Aldrich is the logic of the long coalition. The intense minorities who initiate parties in our model have essentially the same motive as in the original “Why Parties?” story: The opportunity to get a larger share of policy benefits by excluding other interests from the benefits. 16 • Our model focuses on groups in society -- intense minorities, and their opposite, a non-intense and often inattentive majority – as opposed to legislators in Congress. While the intense minorities are close analogs to Schwartz and Aldrich’s legislators, 9 their theory has no counterpart to out uninvolved majority. As a result, many differences follow from the Schwartz-Aldrich story. • In the Schwartz-Aldrich story, the long coalition must be majority, otherwise it has no value. Not so in ours. Owing to voters who may not belong to any intense minority and may sometimes -- we stress, only sometimes -- vote on the basis of empty slogans or lack of understanding of the issues at stake in the election, the party coalition need not represent a majority of interests. Depending on the degree of voter inattentiveness, a party coalition could skew policy in the direction of a rather small set of minority interests. • Owing to their wish to represent less than a majority of interests, parties have an incentive to avoid transparency. Sometimes this may lead to dissimulation, but more often passive strategies suffice to keep the giant asleep. The intense minorities in our founding myth tried to stay in the background in order to avoid decreasing the size of the electoral blind spot. In the Schwartz-Aldrich story, in which parties strive to create genuine majorities, there are no incentives to secrecy or dissimulation. • As in the Aldrich’s account, voters in our story can see what policies the parties publicly support and develop party attachments accordingly. Some attachments may be based on the parties’ meaningless slogans. This is, for both theories, the party-inthe-electorate. • Like us, Aldrich presumes that office-holders enjoy office-holding per se, but this does not create tension between the party and its incumbents. In our account, officeholders may come to value office-holding so much that they ignore party goals and focus exclusively on wooing ordinary voters. • Because of this danger, the party as a coalition of interests needs not only to be concerned with initial nomination and election of candidates, but also with monitoring and sanctioning of their behavior in office to keep them loyal to party goals. These differences with the Schwartz-Aldrich formulation are, for the most part, not empirical; they are more about theoretical interpretation of facts. For example, Aldrich explains in much detail in Why Parties? how party activists sometimes pressure officeholders to take extreme positions. In his theoretical accounting, these activists appear to be extra-party actors, pressuring the office-holders in primaries and by other means to adopt policies the politicians would on their own regard as electorally risky. We recognize the same facts, but view the activists as engaging in an intra-party exercise to 9 Aldrich’s book contains an extensive treatment of activists who closely resemble our intense minorities, but they are outside his basic conception of what a party is. For us, the coalition of intense minorities is the party. 17 define the nature of the party coalition. We recognize that office-holders have somewhat different goals than other party members, but see them nonetheless as all part of the party. Our notion of party bears some similarity to that of E. E. Schattschneider in Party Government (1942). Schattachneider defined party as “an organized attempt to get control of government.” Those struggling to get control prominently included intense minorities — or pressure groups, in Schattschneider’s lingo — as well as local machine politicians, whom our model would treat as extensions of the candidate (because of the value that they place on winning per se.). Schattschneider’s party is thus located “out in the country” rather than in government, 10 and it is far from being a majoritarian institution. In one famous passage, Schattscheider described the election of the president of a boys club. Each boy voted for himself, except two, who voted together and thereby got control of the club. In one sense, the logic here is extremely close the Schwartz-Aldrich model: organized coalitions beat fragmented individuals. Yet, the outcome in Schattschneider’s election is notably non-majoritarian. The example highlights one reason why the question of whether parties originate the in legislature or “out in the country” matters. Legislative decisions require majorities, and a party that originates there will be shaped by this fact. In the US, electoral decisions require only a plurality. The difference between majority and plurality rule is probably not as important as the difference made by the presence of the large number of uninformed voters “out in the country.” But the two differences reinforce each other. A long coalition formed out in the country need not try to benefit a majority, it need not represent a majority. Elections need not be won by the party whose platform is better for more voters, or (as our founding myth showed) even by the party that represents a majority of the organized. In Schattschneider’s theory, parties are not simply the sum of their interest group demands. As in our founding myth, the groups must bargain out a “united front” that is acceptable to all of them. At one point in Party Government, however, Schattschneider declared that “If the parties exercised the power to govern effectively, they would shut out the pressure groups” (192). This extreme statement runs counter to our argument. We agree with Schattschneider that parties operating as coalitions discipline groups; the bargaining process through which parties arise requires groups to scale back their demands. But we would interpret a government that truly shut out its interest groups as a incumbent who had shut out parties. In the comparative assessment of our theory and that of Schwartz-Aldrich, the questions are which does a better job of organizing the facts of party life, of providing an intuitively satisfying account of them, and of revealing what is important to understand about parties. The central posture of our theory -- the idea that is most important for organizing, explaining, and exposing party behavior -- is that parties are a coalition of intense 10 The quotation marks are ours rather than Schattschneider’s. 18 minorities who provide service to democracy only as a by-product of maximizing policy benefits for themselves. The most important task for students of parties is to understand how this motive affects party behavior and how more specifically the party coalition forms and changes, chooses candidates, and keeps them loyal to party goals in the face of non-support by voters. The comparable posture for Aldrich is that party institutions and rules are the endogenous creation of ambitious office-holders. The core task for Aldrich would be to understand how, over a variety of electoral and historical contexts, office holders manipulate party institutions so as to maximize their personal goals of good public policy, career stability, and power in government. Partisan activists and interest groups enter his story as constraints on office-holders in the achievement of their goals rather than, as in our story, the party itself. To put the question even more simply: Is it more organizationally useful and intellectually satisfying to conceptualize parties as the creature of politicians (who must keep their group constituencies satisfied) or as coalitions of intense minorities (who use politicians as their agents)? EMPIRICAL IMPLICATIONS OF THE THEORY Our theory has been constructed to explain diverse and important empirical regularities associated with party politics. We shall now present evidence of these regularities and explain how the theory accounts for them. Candidate positions in Congressional elections A central claim of our model is that parties can take positions away from the position of the “median voter” because voters are insufficiently attentive to electorally punish them for doing so. Voter information is not, however, constant. In this section, we suggest that increases in voter information are, as expected by our theory, associated with greater party responsiveness to voter opinion. A. Effects of media coverage on candidate positions in House elections. In a study of the effect of media coverage of MCs, Cohen, Noel, and Zaller (2004) present evidence that Members of Congress (MCs) establish roll call records closer to median district opinion in their district when media coverage is high rather than low. Figures 2 and 3 on the following pages give the flavor of the analysis. The first figure shows the relationship between presidential vote at the district level (scored as percent Republican) and the NOMINATE score of the MC from that district (with conservative voting records scored high). The four panels of the figure show this relationship for all 435 congressional districts at four points in time. As expected, the relationship is positive in all four time periods. 19 Yet more striking than these positive relationships is the change that occurs over time: Not only does the positive association between presidential partisanship and congressional vote record become stronger, but, more strikingly, district voting becomes more polarized. Especially in 2002, one finds almost no districts with middle-of-the-road MCs (that is, MCs with NOMINATE scores around 0). Even districts that are 50-50 in their presidential partisanship produce MCs that vote either markedly liberal or markedly conservative, with few MCs having a middle-of-the-road voting record. We shall consider the reasons for the change over time later on. At this point, let us concentrate on 1994, the last year of Democratic control of the House. For that year, Douglas Arnold measured the amount of newspaper coverage of the MC in a non-random but large fraction of House districts. For these districts, Cohen, Noel, and Zaller examined the relationship between presidential partisanship and MC voting for those districts in which media coverage of MCs was heavy and in which it was light. Most of the moderate MCs, as it turned out, came from districts where they got heavy media coverage; extreme MCs tended to come from districts with lighter to virtually nonexistent media coverage -- and to be roughly equally extreme regardless of district partisanship. Thus, the relationship between district partisanship and MC voting record was strikingly different, depending on whether voters were supplied with information about the MC. Figure 2 summarizes the pattern. In districts in which the information supply was heavy, MCs became steadily more conservative as their district became more Republican in its presidential voting. But in districts in which the information supply was minimal, MCs were extremely liberal up to the 50 percent threshold in the presidential vote; above that point, MCs suddenly became extremely conservative, with scarcely anyone in the middle. The former were, of course, Democrats and the latter were Republicans. The conclusion the Cohen, Noel, and Zaller draw is that MCs will be responsive to median opinion in their districts only insofar as voters can learn about their records (and challenger campaigns) in the local media. 11 Because Arnold’s data involved a non-random sample of newspapers, Cohen, Noel, and Zaller (2004) extended the analysis to all 435 districts. The level of media coverage was inferred by the degree of overlap between the congressional district and local media markets. Congruence was defined in terms of overlap. If one media market served exactly one congressional district, congruence was high -- and voters presumably got a good deal of information about their MC as a result. But if congressional districts were splayed across many media markets, or if one media market contains 10 or 20 congressional districts, congruence is low -- and voters presumably got little information about their MC because no media outlet had much incentive to provide it. Even candidate advertising would be inefficient and hence low in non-congruent markets. The expectation was that, as in Figure 2, MCs would hew closer to median opinion when congruence was high than when it was low. This pattern held, albeit more weakly for the weaker measure, over the entire period 1972 to 2002. Arnold measures coverage over the MC’s entire term, but this paper uses coverage only from the first 18 months of the term so as to avoid endogeneity bias. 11 20 What probably happened in the light media districts was that voters entered the voting booth with virtually no information about their MC except her party affiliation, with the result that underlying district partisanship almost entirely determined the vote. If an extreme MC were challenged by a moderate of the other party, voters had no way to find out about the challenger’s moderation. This frees MC’s to ignore ordinary voters and instead serve the more ideologically extreme interests of the party. In the heavy media districts, by contrast, extreme MCs could be beaten by moderate challengers because information about the opposing positions was available to voters. The major caveat on these results is that the number of districts in which media coverage is heavy is small -- perhaps 5 percent of all districts. The modal pattern is light coverage in districts in which there is little congruence between media market and congressional boundaries. Thus, variation in media coverage explains relatively little variance in the extremism of MC roll call records. But the theoretical point nonetheless holds: More informed voting, even if unusual, is associated with greater responsiveness to district opinion. B. Historical trends in candidate positions in House elections. As just shown, variation in the information supplied to voters explains cross-sectional variation in the responsiveness to district opinion. But what can explain the substantial over-time variation that is also clearly present? To answer this question, we turn to the longer time series constructed by Ansolabehere, Stewart, and Snyder in their paper, “Candidate Positions in Congressional Elections.” The main finding in that paper is that, through most but not all of the period from 1874 to 1996, MCs’ roll call voting did not converge to the median of opinion in their district. As in the Cohen, Noel, and Zaller study, MC positions are measured by NOMINATE scores and district partisanship is measured by the presidential vote. This basic finding -- that Members of Congress frequently fail to represent median opinion in their district -- is consistent with our theoretical model. But what of the exceptional periods, in which MCs were more responsive to median opinion in their districts? Ansolabehere et al. are able to identify these periods with some exactness. Interestingly, they vary by party. After six decades of non-convergence, Republican MCs began to converge to the median position of their districts in 1934, became gradually more convergent until about 1970, and then began to diverge through the early 1990s, whereupon the pre-1934 level of (non) responsiveness had re-emerged. On the Democratic side, MCs remained more liberal than their districts through the mid1960s, whereupon they became suddenly more responsive to median district opinion. But by the early 1990s, Democratic were again making voting records that were more extreme than the districts they represented. Our explanation begins with the observation that salient new issues emerged almost exactly at the time when parties began to converge to median opinion and, further, the parties that converged were the parties that were thrown on the defensive by the new issue. In 1934, Republicans were put on the defensive by the popularity of Roosevelt’s 21 new big government liberalism. In the late 1960s, the Democrats were put on the defensive by a cultural backlash against the 1960s, perhaps especially the spread of civil rights enforcement throughout the nation. The issue was sufficiently salient that Democratic candidates had to respect voter opinion on it. Consistent with our model, we suggest that voters, though rarely very attentive to politics, are sufficiently attentive that MCs must defer to them on highly salient, politically charged new issues. Voters, as Key says, are not fools; they are merely inattentive much of the time. If, for some reason, they do become attentive, parties must respond. A party that finds itself on the wrong side of history is a party that has drifted out of the electoral blindspot. By definition, a salient issue is one that voters attend to with unusual care. As Figure 4 illustrate, the extra attention need not be symmetric. In Figure 4, voters are more sensitive to policies that they perceive as too liberal than to those they perceive as too conservative. As before, we assume that voters do not notice many policy differences, so that indifference curves are thick. The smaller lopsided circular curve denotes the electoral blindspot – parties can pick candidates with positions in this area with impunity. The shape of the electoral blindspot in Figure 4 is intended to capture our conjectures about effect of the counter-culture backlash on the Democrats in the 1960’s. The electoral blindspot is flattened on the bottom, indicating that the voters are more likely to notice and react to positions that too socially liberal than to those that are too socially conservative (or too extreme in either direction on the economic dimension.) Both parties want to pick candidates on the edge of the electoral blindspot, the Democrats in the lower left quadrant, the Republicans in the lower right. The extra salience of too much social change, too fast, forces the Democrats to move closer to the voters than the Republicans. The important point in Figure 4 is that even though the Democratic candidate is much closer to the voter’s ideal point than the Republican, the Democrat’s chance of winning is no higher. Both candidate’s positions are perceived as equivalent by the partly-attentive majority. Something other than policy position will be decisive in this election. More investigation is obviously needed before this explanation can be accepted as conclusive. In the meantime, we advance it as plausible. And we continue to claim the main finding of Ansolabehere et al. -- that MCs diverge from median opinion in most districts most of the time -- as consistent with our theoretical argument. Along with our earlier discussion of the effect of media coverage on candidate positions, we take this as supportive of the theoretical claim that parties wish to exploit shortcomings in voter information to elect candidates who are more extreme than voters would prefer if better informed. 22 Figure 4: Voters More Attentive to Social Liberalism that Social Conservatism. Social Conservatism R V Economic Conservatism D Candidate positions in Presidential elections These results highlight what is, in our model, a central problem for parties, namely, to estimate how close their candidates must converge to the middle of the ideological spectrum in order to win elections. Because voters often chose on the basis of random factors, such as how the economy has performed in the last few months before the election, and because the pivotal voter has some but not great ideological acuity, parties have slack. But how much? The figure below suggests that, even in high profile presidential races, the amount of slack is considerable. The histogram (or bar graph) shows the public’s responses to a survey question asking “Generally speaking, do you consider yourself a liberal, a conservative, or what.” Follow-up questions ask whether people consider themselves slightly or extremely liberal or conservative. As can be seen, a large majority of voters are either in the ideological center, “slightly” ideological, or unaware of their ideology. 23 Yet, when survey respondents were asked to rate the ideological location of the candidates, they generally rated them further from the center -- as straight-out liberal or straight-out conservative. If the parties saw advantage to nominating centrist candidates, they could certainly do better than they do. If, as we maintain, the goal of parties is to nominate the most extreme candidate having a good chance to win, parties should test the limits of voter tolerance. They should, in other words, occasionally nominate a candidate who is perhaps too extreme to win in order to be sure that the candidate really cannot win. Then, if the candidate does fail to win, parties should back off and nominate a more moderate standard-bearer in the next election. The following figure shows that parties do this. The figure shows the ideological distance of presidential challengers from the ideological center in each election from 1948 to 2000. The units of ideological distance are the same as used in the previous graph. 12 As can be seen, the more consecutive presidential elections a party has lost, the more likely the party is to nominate a candidate who is close to the ideological center. Challengers never reach or even approach the ideological center, but electoral defeat impels them to move somewhat closer. Challenger positioning in presidential elections 2.25 McGovern Goldwater Reagan 2 Distance 1.75 from Ideological Center 1.5 Dole Mondale Dukakis Bush II Carter Kennedy Nixon Dewey Clinton Ike Stevenson 1.25 1 1 2 3 4 Number of terms without party control of presidency 5 It may be useful at this point to re-iterate a point from our earlier theoretical discussion: Although our analysis is based on left-right policy differences, we expect that it applies The ratings of candidates from 1972 to 2000 are the same as in the previous figure. Ratings for candidates from 1948 to 1968 are derived from the expert ratings collected by Steven Rosenstone for his book, Forecasting Presidential Elections. 12 24 equally well to non-ideological matters. In every area of political disagreement -campaign finance reform, anti-corruption measures, tariffs -- parties will want candidates to take the position most favorable to its groups and activists that they can take without unduly irritating pivotal voting blocks. The aim of parties is not to give voters what they want on such issues, but to give them as little of what they want and still win. The relationship between party and office-holders A. Introduction. Primary elections are open to ordinary voters but are specified in our theory to be low information contests. By coordinating behind and working for a single candidate, party interest groups and activists can normally prevail in these elections. In the bare theoretical world we have sketched, it seems likely that nominees would emerge from the community of party activists and interest groups and would be, for this reason, likely to share their values and policy preferences. We therefore consider officeholders to be bona fide members of the parties in whose names they run. At the same time, office-holders experience an incentive that other party players do not -the incentive to hold on to office for its own sake. This can induce them to pander to voter opinion -- that is, to cede more to voters than is absolutely necessary to be reelected -- in order to increase the chances of re-election. This is an important tension in our model: Parties know that office-holders must cede some policy to voters, but they would like it to be as little as possible so that party interests can enjoy as much as possible. Parties must thus balance opposing risks -- the risk of extracting less policy than they might ideally get, versus the risk of losing the election and extracting no policy benefit at all. These risks are in turn affected by many unknown quantities, including the concessions the other candidate may make, political events that may influence election outcomes, and the dynamics and serendipity of political campaigns. Office-holders and candidates may trade off these risks differently than interest groups and activists because the former value office-holding per se more than do the latter. The analogue in capitalism would be a company salesman who, in a bid to increase sales and please his customers as much as possible, offered goods at no profit. Cutting profits would increase sales and customer satisfaction. But since it is profit rather than sales that companies care about, such behavior would displease the company. It is the same with parties. They want their candidates to please voters and win elections, but pleasing voters is a means rather than an end. The end is policy that is congenial to the party’s minority interests, whether voters want it or not. The company salesman who enjoys sales more than profits is not much of a problem for business. This could be because the pleasures of sales per se are not very great, but it could as well be explained by the ease of monitoring a sales staff. Neither condition holds as strongly in politics. Office-holding is an intrinsic pleasure and monitoring of office-holders by their principals is, as we show below, often quite difficult. 25 Parties thus find face a somewhat loose set of market constraints. Voter ignorance sometimes permits them to achieve policy benefits that are much larger than a more informed electorate would permit. And yet parties can also sometimes lose control of their agent office-holders and reap little or no gain. In the last section, our focus was on the first of these constraints – the pull of voters, especially centrist voters, on parties and party office-holders. Information, as we saw, helped voters get more of their way. In this section, we shall deal with the second of the constraints – the pull of the main body of party activists and interests on office-holders. As shall see, various forms of information will be a critical determinant of who gets their way. B. The role of primary elections. Many authors have noted that politicians must navigate between the pressures of two electorates, a primary electorate and a general electorate. Our theoretical account is a bit different than some in maintaining that 1) it is not the primary electorate but the party activists who are the source of centrifugal pressure and 2) the party activists behave strategically to pull the candidates away from the center to an optimal degree but not arbitrarily far, and 3) voter ignorance and inattentiveness aids activists in their efforts. That primaries per se are not the source of centrifugal pressure has been demonstrated by Ansolabehere, Stewart, and Snyder. Since, as they point out, MCs remained more extreme than district opinion even in the 19th century, which was before primaries existed, primaries cannot explain the phenomenon. Our theory holds that party activists and interests control nominations through primaries. They do so by coordinating electoral resources behind a good representative of party interests, thereby enabling that person to prevail in what is normally a low-key election. Some recent studies provide evidence for this view (Cohen et al, 2001, Masket, 2004). If necessary, activists can also use the primary to sanction office-holders, threatening them with removal for poor performance in office. The main reason sanction might be necessary is that office-holders might be tempted to cede too much policy to voters. Even though primaries were often instituted by Progressive reformers to break the control of parties over nominations, there is much evidence that parties can in fact control primaries. Indeed, some state party organizations actually promoted primaries. Writing about the effect of changes in nomination procedures 1909, H. J. Ford observed: One continually hears the declaration that the direct primary will take power from the politicians and give it to the people. This is pure nonsense. Politics has been, is, and always will be carried on by politicians, just as art is carried on by artists, engineering by engineers, business by businessmen. All that the direct primary, or any other political reform, can do is affect the character of the politicians by altering the conditions that govern political activity, 26 thus determining its extent and quality. The direct primary may take advantage and opportunity from one set of politicians and confer them upon another set, but politicians there will always be so long as there is politics. (cited in Key) Still, primaries present some challenge to parties. Defeating an incumbent in a primary requires finding a strong challenger and mounting a campaign in a low-key race that is likely to afford a great natural advantage to the incumbent. It is therefore worth testing whether parties can always sanction office-holders for unnecessarily moderate behavior. Using the media congruence data described above, the following figure tests the likelihood that a “moderate” incumbent can be defeated in his primary at different levels of media congruence. The idea behind the test is that sanctions will be more difficult where media congruence is weakest, because no one will be able to effectively challenge the incumbent. As can be seen, this is what occurs. The ability of a moderate incumbent to defy his party activists with impunity is greatest when market congruence is lowest. Probability that moderate MC (Nominate =0) will be defeated in primary 12% Safe seat for party Probability of primary defeat (each year) 8% Marginal seat for party 4% Seat in other party territory 0% 0 0.1 0.2 0.3 0.4 0.5 Media market congruence This argument can be usefully expressed in principal-agent terms. An office-holder is the agent of two principals -- the party that nominates him and the voters that elect him. Information, measured here as media market congruence, increases the ability of both principals to monitor. In the general election, it empowers the pivotal median voter, enabling him to defeat extreme incumbents in swing districts. Absent information, the agent (office-holder) shirks on this principal and serves the other, the party that nominated him. However even the party may have enforcement problems. When media congruence is low, the party cannot extract the extreme agent behavior it would like, even in districts that are safely one-party. These results, we should add, do not mean that party influence is nil in low media districts, because party activists and interests can still presumably control open primaries. 27 The capacity to pick, or at least to influence, the person who becomes the incumbent in the first place is extremely important even if that individual is free from subsequent sanction. Yet the question arises, what if the party were entirely frozen out of the nomination process? What if it could not even control the selection of nominees in open primaries? In this rather extreme scenario, one could no longer be clear that the person who ultimately won office would be a party member. He might be almost any type of person - the agent of a particular interest group rather than a coalition of interests, the agent of general election voters alone, or perhaps most likely, an independent entrepreneur. Seth Masket’s (2004) study of California legislators is consistent with these conjectures. When, as he shows, an unusually draconian set of Progressive reforms effectively severed office-holders from party activists, legislators lost interest in partisan organization of the legislature. They chose the Assembly speaker on a non-partisan basis and did not form strongly partisan voting coalitions, as measured by NOMINATE. At one point in the 1930s, a Democratic majority elected a Democratic speaker with a view to using party discipline in service of a liberal agenda. However, the legislature revolted and the speaker was replaced within a few months. Legislators continued to appeal to party labels in elections, but were unwilling to accept any sort of party program. Interest groups outside the California legislature – labor unions on the liberal side and the fledgling California Republican Assembly on the conservative side – grew frustrated by this situation and agitated to end cross-filing, which was the principal means by which legislators avoided control of activists in their districts. (Cross-filing allowed incumbents to run in both party primaries without revealing their party affiliation. This gave officeholders themselves virtually total control over their re-nomination, insulating them from monitoring of party activists.) Legislators resisted the end of cross-filing, but interest groups “out in the community” sponsored a voter initiative that ended it anyway. Almost immediately, the majority Democratic Party in the legislature began, under the leadership of Jesse Unruh, to vote in a more disciplined fashion and to pass important liberal legislation. Legislators complained about the new partisanship, and elements of the old fake partisanship remained for some years, but Masket identifies the repeal of cross-filing as the turning point. The effect of cross-filing, as Masket argues, comports poorly with the Schwartz-Aldrich story of party origins. In that story, legislators take the initiative in forming a long party coalition, and they do so because they value the greater policy benefits that it can produce. But when, as Masket finds, office-holders become independent of activists, they may cease to be partisan. This suggests that the impetus for partisan organization comes from interest groups and activists “out in the country” rather than from the officeholders themselves. 28 As Aldrich has added, long coalitions become parties in part because they are useful to office-holders in their efforts to be re-elected. Our theoretical account is different. The coalition of party interests does provide candidates with valuable electoral support, but it also asks candidates to take positions that entail more electoral risk than the candidates would prefer. Even more importantly, the formation of the first partisan coalition will induce the formation of an opposing one, which incumbents will experience as an institutionalized source of electoral opposition. Long, highly partisan coalitions are good for the minority interests and activists that originate them, but they are not necessarily good for office-holders intent on a long career in office. This is why office-holders may shirk from party discipline when able to do so. In Why Parties? Aldrich provides a compelling account of why party activists are likely to pressure party candidates to become more extreme than office-holders wish to be. We not only agree with this argument but have incorporated much of it into our theory. So we are certainly not saying that Aldrich is unaware of activists who pressure officeholders and candidates to become more extreme. The difference is one of overall theoretical accounting. As we understand Aldrich’s overall argument, he views the coalition of office-holders (and the electoral institutions they create) as the party itself, and he views activists as apart from or potentially opposed to the party and its interests. When, in Aldrich’s theory, activists pressure candidates to move away from the ideological center, they do so from positions outside the party. In our theory, however, the activists are as much a part of the inner constituency of party as are the candidates and office-holders. The essential activity of these various actors, as we have suggested, is to figure out how to form a united front that can unite the party and still win elections. Thus, activists and interest groups that exert pressure to move away from the center are not engaging in action against the party; they are party members who are engaged in making a central and difficult calculation. As we have noted, candidates and activists do tend to have different perspectives -- office-holders want to cede more to voters than activists -- but it is by no means obvious that candidates are always more correct in their judgment of optimal location than are the activists. The Polarization of Congress from the 1980s to the present. In his classic book, The Electoral Connection, David Mayhew explains how the logic of electoral competition played out in the Congress of the 1960s and early 70s. MCs, as he argues, would appeal to partisan interests at election time when it would help them with voters in their districts, but they had no use for party discipline in legislating. They organized Congress so as to maximize their freedom for credit-claiming and serving local constituency interests. The most precious commodity that Congress controlled was money for pork-barrel spending, which was distributed on a universalist rather than partisan basis, so as to help all incumbents. This practice guaranteed that, in the event of a change of party control, all incumbents would continue to have the wherewithal to assure their re-elections. In this way, MCs were a conspiracy of the elected against the 29 unelected. A later paper by Weingast (1979) provides formal justification for Mayhew’s general view. As many scholars have noticed, this behavior began to change in the 1970s and is dramatically different today. Party, and party discipline, have become as important in the contemporary Congress as they were unimportant in the one Mayhew described. Our theory provides a ready hypothesis to explain this change. To return to the language of principal-agent theory, the capacity of party principals to monitor and control the behavior of MCs must have increased relative to that of the median voter. We are not in a position to systematically test this hypothesis, but, at an aggregate level, we can cite important evidence that fits it. Over the period since the 1960s, there has been a dramatic increase in the number of interest groups active in national politics, including ideological advocacy groups. These groups not only lobby MCs, but fund and monitor them. Many groups sponsor report cards and newsletters. The ease of transportation and communication and, most recently, the Internet has further increased the ability of interested groups to monitor the behavior of MCs. Mayhew’s theory of the electoral connection, with its emphasis on credit-claiming, position-taking, and failure to deliver partisan policy, is essentially a theory of bamboozlement. MCs would take partisan stands but not partisan actions and no one outside of the chamber itself could tell the difference. Today’s interest group and ideological activists know the difference very well. Today’s party coalitions are also more ideologically congenial, as southern, conservative Democrats have moved into the Republican Party. How that came about is an important chapter in party history, which we shall discuss below. 13 As we shall discuss in the next section, the internal organization of Congress has also contributed to improved monitoring, but, as scholars recognize, that development must be seen as the endogenous effect of some other cause. That cause, as we hypothesize, is exogenous improvement in the ability of party interests and activists to monitor their MCs. Viewed over a longer time frame, the ability of parties to monitor MCs must have been at something of a 150-year low point in the late 1960s. Urban machines, in which professional politicians monitored professional politicians, had been in decline for several decades and were on the verge of extinction at that point (Mayhew). Another important blow hit just at that time: Up until the Supreme Court’s one-man, one-vote forced dramatic reapportionments across the country, party organizations were typically organized along the boundaries of counties. But the requirement for equal-sized districts in the 1960s destroyed that system almost overnight, severing many MCs from the organizations that had nourished them (Katz and Cox, 1996). Beginning in the 1970s, a trend toward better monitoring set in, as just described 13 See for instance Koger, Masket and Noel (2006) for an attempt to measure these ties. 30 Obviously, much more research is necessary to establish this argument. But it is, we believe, a theoretically cogent and empirically plausible one. Even in its provisional state, the argument accomplishes a useful theoretical purpose: To explain how the logic of Mayhew’s Electoral Connection and that of the Schwartz-Aldrich “Why Parties” story can both be correct. The key to that resolution is to recognize that the long coalition forms outside rather than inside of government, and that variation in the ability of interests and activists in the long coalition to monitor their agents inside of government can explain when legislators behave like Mayhew’s MCs and when they behave like those of Schwartz-Aldrich. Where, in practice, does the long coalition form? The notion that the long coalition forms outside of government is thus a fundamental part of our theory. 14 Many theoretical assumptions cannot be directly tested; the researcher can test only their implications. In this case, however, something like direct testing is possible. Aldrich offers direct evidence on this point. He notes that in the first sessions of the American Congress, legislators did not coalesce in stable partisan blocks, often even falling into cycles in which one bill would defeat a second which would defeat a third which would then beat the first. Aldrich attributes the formation of parties to a pact among leaders in Washington to form, in effect, a long coalition. Aldrich’s account is plausible but not definitive. When the long coalition formed, it formed on the basis of what Aldrich calls “the great principle,” which was the longstanding disagreement between Federalists and anti-Federalists. Since it pre-dated the Congress, this principle represented a coalition that had somehow formed “out in the country,” perhaps in response to debates over ratification of the Constitution if not before. Further, the great principle began to structure voting in Congress at about the same time that “democratic clubs” of what eventually became the Jeffersonian Republican party were forming and pressuring Congress. It does seem to be true that the first members of Congress were not elected by anything resembling parties; it is not as clear that partisanship was wholly the idea of the office-holders rather than of the people who had sent them there. The formation of the first presidential parties is similarly ambiguous. By the election of 1800, the Federalist and Republican parties existed both in name and in fact, and both held caucuses among their congressional delegations in Washington to nominate a president. The dominant Democratic-Republican Party continued to hold caucuses among its MCs in Washington until 1824 and, until the final year, its nominees always won the presidency. It is reasonable to interpret this activity as parties that originated in Congress and moved outward to colonize other parts of government. It would seem, that is, to fit the Why Parties? story. By “outside the government,” we do not mean to exclude anyone who holds office – federal, state or local. Rathter, the point is to include all of those office holders, as well as those who do not hold office. The party forms in the country as a whole, and not merely within the federal legislature. 14 31 The Federalists, however, had so little strength in Congress that they began to meet instead in New York. As Gerald Pomper describes the Federalist nominations: The party’s strength in the national legislature was too small to be representative of the group. Nominations were made basically by small groups of party leaders. In 1808 and 1812, the choice was made at primitive national conventions. The one in 1808 was essentially the first in American politics, though it bears little relationship to the present system. One half of the states were represented at this meeting, which was closed to persons not specifically invited and to the general public. Similar procedures were followed in 1812. In the following election, the last contested by the Federalists, Rufus King was named by the party without any formal action (1966, p. 9). The change of venue and procedure has significance. The Federalist Party nomination procedures, and especially the notion that Federalist legislators were not representative of the Federalist Party as a whole, do not fit a theory of legislaturecentered parties. They suggest, rather, a theory in which party division may originate “out in the country” rather than in Washington and in which partisans try, even if not always successfully, to send politicians to Washington who will represent their views. . Having lost five straight presidential elections, the Federalist party declined to make a nomination for the presidential election of 1820. This allowed James Monroe to cruise to re-election in a suddenly non-partisan “Era of Good Feelings.” But feelings were not all good within the ranks of the victorious party. One problem was patronage. When the Federalist Party existed, Republican presidents Jefferson, Madison, and Monroe gave fellow partisans priority for federal jobs. But with the demise of competition, Monroe began appointing former Federalists to plum positions. This led to shrieks of protest from Martin Van Buren, the leader of the New York state Republicans and an ardent advocate of patronage to cement party discipline. Van Buren’s first act when he came to Washington in 1820 as a New York senator was to challenge Monroe’s universalistic patronage policy. The challenge was not successful, but Van Buren’s effort drew praise from Republican editors around the country. The other problem was presidential succession. As Monroe’s retirement approached, at least seven men, all claiming to be good Jeffersonians, sought a nomination that could be tantamount to election. The caucus system could not contain this conflict and it immediately came under an attack that ended its authority. Four candidates remained active, with the result than no clear victor emerged.. As might have been expected, the candidate who was unopposed in his region, John Quincy Adams, won election, but he needed a vote in the House of Representatives because no candidate gained a majority in the Electoral College. This outcome was distressing to supporters of Andrew Jackson of Tennessee. He won 41 percent of the popular vote compared to only 30 percent for Adams, and he polled a narrow plurality in the Electoral College as well. Soon after the election, supporters of Jackson organized a primitive front to work for the Tennessee hero and to make sure that he had no important competitors for the votes of his natural constituents. 32 The leading architect of this front was Van Buren. His plan was to bring his state party in New York State into an alliance with state parties in the South that would dominate national politics. The advantage to New York would be a reliable flow of federal patronage. The advantage to the South would be the protection of slavery. As Van Buren wrote to James Ritchie, the leader of what was called the Richmond Junto in Virginia: We must always have party distinctions …. If the old ones are suppressed, geographical division founded on local interests, or what is worse prejudices between free and slave holding states will inevitably take their place… If, as Van Buren thus argued, southern parties failed to ally with a big northern state in a new national party, a party system pitting north against south would likely form in its stead, thereby bringing slavery to the fore of national politics. Van Buren made a famous tour of the South in which he pressed his argument on other regional party leaders. He also set up committees to coordinate electoral activities in other parts of the country and raised money to start pro-Jackson newspapers. By these means, Van Buren cobbled together an alliance of northern and southern state parties under the banner of a new national party and got them to turn out their voters in support of Jackson. Jackson himself played little part in the formation of Van Buren’s new party; in fact, Van Buren was unsure whether Jackson, who has making electoral efforts on his own behalf, would even go along with it. But “Old Hickory” did and the result was victory. The new alliance won four of the five presidential elections beginning in 1828, including one with Van Buren as its nominee. Van Buren’s victorious alliance set up permanent shop as the Democratic Party. As in Aldrich’s account of party formation, the new entity was a long-term logroll between different groups, each getting what it wanted through the support of its coalition partners. New York got the rich flow of patronage that Van Buren had long coveted. The Democratic Party’s slogan, that “To the victor belongs the spoils,” was in effect an expression of the modern idea of a “minimum winning coalition” in the division of government benefits. The white leaders of the South made out even better. For them the key principle was, in effect, “Who nominates the candidates controls the national agenda.” As long as the new Democratic Party consisted of an alliance of northern and southern states, it could not nominate a presidential candidate who would make an issue of slavery. It is notable that the new system originated in state party organizations rather than in Congress – this time more clearly “out in the country” than in Washington. Van Buren was, to be sure, a U.S. senator, and he used his Washington position to agitate for Jackson and against Adams. But in stitching together the coalition of “Jackson men” he was not working with other senators within the Senate, but with political leaders across the country 15 Of course, legislative politics can, as Aldrich’s account emphasizes, also build “long coalitions” that deliver patronage and agenda control. But, in this instance, legislative 15 This is the historical consensus. But we need to read more. 33 politics were not doing so — at least not to the satisfaction of locally based interests. Hence local politicians, based on a clearly stated view of state party interests, took matters into their own hands. We stress the group origins of the new party system because, as noted earlier, theorists of political parties often view them as originating in legislative politics. The foundation of the Democratic Party demonstrates an alternative possibility. The Whig Party also did not come together in the legislature, mainly because Clay, Webster, and Calhoun were all in the legislature in the mid-1830s as it was forming and were all running for president. When the first Whig convention met, in response to a call from state legislatures, it nominated none of the above, an aging Indian fighter who was bland enough to unite a party of minorities. The Harrison coalition was put together at the nominating convention, not in the Congress. Interestingly, Max Weber described party development in the United States very much along the lines just sketched: In the beginning, when parties began to organize, the members of the House of Representatives claimed to be leaders, just as in England at the time when notables ruled. The party organization was quite loose and continued to be until 1824. In some communities, where modern development first took place, the party machine was in the making even before the eighteen-twenties. But when Andrew Jackson was first elected President--the election of the western farmers' candidate--the old traditions were overthrown. Formal party leadership by leading members of Congress came to an end soon after 1840, when the great parliamentarians, Calhoun and Webster, retired from political life because Congress had lost almost all of its power to the party machine in the open country. That the plebiscitarian ‘machine’ has developed so early in America is due to the fact that there, and there alone, the executive--this is what mattered--the chief of office-patronage, was a President elected by plebiscite. (Politics as a Vocation, pp. 107-108) The American party system retained its intensely local character through much of the country’s history. Describing national party organization as it appeared in 1965 – or shortly before the reforms that created the new primary-based system – V. O. Key, Jr. wrote: Viewed over the entire nation, the party organization constitutes no disciplined army. It consists rather of many state and local points of power, each with its own local following and each comparatively independent of external control. Each of the dispersed clusters of party professionals has its own concerns with state and local nominations and elections. Each has a base of existence independent of national politics. Each in fact enjoys such independence that more than a tinge of truth colors the observation that there are no national parties, only state and local parties. (1965, p. 329) 34 Thus, our national party system has never been an imposing edifice. Surveying the political landscape, one might even miss it. “The closer one gets to our two great national political parties, the more difficult it is to find them,” said political scientist Stephen K. Bailey of the old system. 16 And yet, like the caucus system that preceded it, the multiple independent actors have often been capable of pulling together, uniting behind a national candidate, and getting a large fraction of the electorate to vote for that candidate. Only with difficulty can it be maintained that this party system originated in legislative politics. Change in the party coalitions. Over the course of American history, the party system has, of course, changed a great deal. Some change has been due to changes in the policy demands of core groups in one party or the other (Karol, 2005). Some, however, has been due to the entry of new groups into the coalition or to the rearrangement of these groups across parties. The latter are among the most important in politics, and they are what concern us here. If the party coalitions are worked out and maintained primarily as legislative deals, we might expect to observe the legislature at the center of these realignments. Or, likewise, if parties are the creatures of office-holders, we should find that office-holders lead the rearrangement of party coalitions. If, on the other hand, party coalitions are assembled, as we maintained, groups and activists “out in the country” and merely enforced on the legislature and on office-holders, realignment should appear first in party organization — and, in particular, in the nominations that parties make — and only later in the politics of the national legislature. There is, of course, a vast literature on political realignments, much of it controversial or discredited (see Mayhew). Essentially none of this literature, however, has ever found the original source of realignment in Congressional politics or in the efforts of officeholders in general. Congressional politics and the changing positions of office-holders are much more often the lag indicator, responding either to electoral upheavals or presidential initiatives that are taken to be the actual source of realignment (e.g., references). At the same time, the role of groups and activists “out in the country” has also not played a feature role in the realignments literature. Yet, as we shall sketch in this section, there is ample evidence the role of groups and activists has been extremely important on several occasions, and — more critically to our theoretical argument — that their importance has been centered in their ability to control nomination to office. A. The formation of the Republican party. As is well known, the Republican formed from public outrage in the North against the Kansas-Nebraska Act. Critically for our theoretical account, this new party did not form in Washington. Although it is true that northern Whigs and Democrats opposed the act in Congress, William Gienap’s classic account of The Origins of the Republican Party, 1852-1856 argues that their cooperation was “distinctly limited.” 16 Cited in Broder, 1971, p. 228. 35 ...the bipartisan outcry against the [Kansas-Nebraska act] in the North stood in sharp contrast to the crippling divisions among opponents in Congress. Free Soilers, Whigs, and anti-Nebraska Democrats voted together, but they would not work together. This failure provided scant encouragements to those who looked for the Nebraska issue to realign the party system (p. 79). When the Republican party formed, it was due to calls from activists ”out in the country” to come together in conventions, most famously in Ripon, Wisconsin, to form a new alliance that would nominate candidates with different commitments than those of the old party system. The new party quickly attracted office-holders of both parties who were eager to remain office-holders, but it was not a party put together by office-holders. B. The “system of 1896.” For many decades, leading scholars believed that the bitterly fought presidential contest of 1896 involved a realignment of electoral forces (Schattschneider, Burnham). More recent analyses have cast doubt on that view (Bartels, Mayhew). The Democrats did badly in that election and for several elections thereafter, but the alignment of forces did not change much if at all. One change that did occur — and that gave many close observers the impression that a realignment had occurred — was the nature of the Democratic party nominating coalition. Long split between gold Democrats from the urban east and silver Democrats from the west, the party was captured in 1896 by its silver Democrats. Thus scholars who, like Burnham, looked to voters as the source of realignment might have found more of what they sought if they had attended instead to party nominations. C. Evolution of the New Deal alignment. We note, first of all, that we do not have a theoretical account for the formation of the New Deal alignment. It is best understood as voter response to New Deal policies and rhetoric, which are phenomena presently outside of our model. We may be able to account for it by adding a president to our model, but we have not yet done so. Our theory can, however, provide a reasonable account of the evolution of the New Deal alignment from an economically class-based system to the current system, which is a blend of race, economics, and moral traditionalism. In this paper we can, of course, only sketch this argument. An early step in the transformation of the New Deal alignment occurred at the 1948 Democratic nominating convention. The party was so deeply divided that it was not clear that, as an un-elected president, Harry Truman could even be nominated. He was pressured on the left by the newly formed Americans for Democratic Action to move a civil rights agenda and by Southern members of the party to keep his presidential hands off race. Truman sided with the ADA, which prompted a Dixiecrat walkout to form a new third party. From our perspective, it is notable that the first major public event in the 36 transformation of the party occurred in the context of a group struggle over nomination politics. Another development was the Civil Rights Act of 1957. This was the first major civil rights act since the Civil War era and a precursor of civil rights laws that were to reshape the party alignments. On the surface, this was a legislative initiative. Yet, by both contemporary and historical accounts, the real impetus was Senate Majority Leader Lyndon Johnson’s ambition to become president (Evans and Novak, 1966; Caro, 2002). Johnson knew he could not be nominated for high office by the national Democratic party without demonstrating his commitment to civil rights, and a major civil rights bill was his means of doing so. As a southerner, Johnson was able to persuade other southern senators not to filibuster the bill with the argument that, if he did not win the Democratic nomination in 1960, the much more racially liberal Hubert H. Humphrey surely would. The civil rights bill of 1957 did not do a great deal to improve the conditions of African Americans, nor did it directly affect the party alignment, but it did make Johnson an acceptable vice-presidential candidate for the party in 1960. But more to our theoretical point, it shows how nomination politics could affect legislative politics. Congressional voting in the period from 1940 to 1970 was strongly influenced by the socalled “conservative coalition.” This was a union of Southern Democrats and Northern Republicans that strikingly resembles the contemporary Republican party. Yet, equally strikingly, it had no organizational basis in congressional politics. As in the response to the Kansas-Nebraska Act, the Conservative Coalition was an ad hoc voting coalition. The two major civil rights acts of 1964 and 1965 were another big step in the realignment. They made clear to anti-civil rights southern whites that the Democratic party would no longer protect them from civil rights legislation, whereupon many deserted to the Republican party, at least in presidential elections. But the impetus to this development came not from inside Congress but from the president. And many of the congressional votes came from Republican MCs who had no wish to prevent the Democratic party from establishing itself as the party of racial liberalism. The final nail in the coffin for the New Deal alignment was the abortion issue. Ronald Reagan, leader of the increasingly energetic conservative wing of the Republican party, embraced the pro-life position just as he began his run for the 1976 party nomination. By the mid-1980s, virtually all major presidential candidates in the Republican party were pro-life and virtually all in the Democratic party were pro-choice. In many cases, the correct party position involved major flip-flops from previous positions, often undertaken at the moment of entering the nomination contest (Karol, 2005). Still, voting in Congress lagged. In the most detailed investigation of the politics of the congressional realignment on cultural issues, Cohen (2005) traces the changes to struggles in local Republican party organizations. Beginning in the middle 1970s and picking up steam through the 1980s, religiously conservative activists began pressing Republican party organizations to back pro-life candidates. When, as often occurred, the old guard resisted, religiously motivated activists challenged them in primaries and 37 eventually won a place for themselves in the party. As Cohen argues, the Republican revolution of 1994 would probably not have been sustained except for the success of religious activists in these intra-party battles. We are under no illusion that the evidence sketched in this chapter proves that party realignments originate in the effect of groups and activists on nomination politics. It suffices only to make it plausible that the dynamics envisioned by our model may correspond to the actual dynamics of party politics. Secrecy and information manipulation. Because parties are a coalition against the median voter, they need to obscure what exactly their policies are and to prevent public votes on policies that would be more popular. This is what, according to many studies, the House Rules Committee does. In a recent study, moreover, Alan Rozzi has shown that the Rules Committee is more likely to craft closed rules on legislation that is central to party interests and to the sentiments of party activists, as expressed in national party platforms. All of this makes sense in terms of our theory of party as coalition attempting to gets its way over popular majorities in the electorate. In David Rhode’s theory of Conditional Party Government, members of the majority party in Congress will empower their leaders to structure legislative politics when the majority party is sufficiently homogenous in its policy preferences. This is consistent with our theoretical view as well. Our emphasis, however, would be somewhat different. We expect party government in situations in which party interest groups and activists are able to effectively monitor MC behavior in office. Concluding comments. Although we have not provided anything like definitive empirical support for our theory, there is a good deal more evidence that supports our general view and will be cited in later drafts of the paper. But, as suggested earlier, the main facts of party politics are pretty well established and generally accepted (Krehbiel notwithstanding). The main question is what one wishes to make theoretically of them. We believe, and have attempted to show, that our theory rather easily organizes and explains many of these facts. And we believe that the notion of parties as coalitions of special interests trying to maximize their share of policy benefits also fits the basic behavior of parties quite well. Our remaining task, which we must leave for future work, is to show that the performance of our theory in these matters is superior to that of the Schwartz-Aldrich alternative. Those efforts will focus on the politics of nominations. Nominations do not loom large in the Schwartz-Aldrich notion of office-holder dominated parties, but they are at the center of our theory. Nominations are, as Schattschneider stated, the foremost expression of what party groups and activists want, and de-nomination is, as we have added, the means of enforcing those preferences on wavering office-holders. As the metaphor of the smoke-filled room suggests, the politics of nominations have always been secretive and difficult to study, and they still are. But, as our continuing work aims 38 to show, nominations are the central act of party politics. By bringing them to the forefront of the study of parties, we hope to achieve a better understanding of what parties are and how they contribute to democracy. 39 References Achen, Chris and Larry Bartels 2004. “Musical Chairs: Pocketbook Voting and the Limits of Democratic Accountability.” Paper presented at the Annual Meeting of the American Political Science Association. Chicago Aldrich, John. 1983. “A Downsian Spatial Model with Party Activism.” The American Political Science Review. 77 974-990. Aldrich, John H. 1995. Why Parties? The Origin and Transformation of Party Politics in America Chicago: University of Chicago Press. Ansolebehere, Stephen, James M. Snyder Jr. and Charles Stewart III. 2001. “Candidate Positioning in U.S. House Elections.” American Journal of Political Science. 45 136-159. Arnold, R. Douglas 2004. Congress, the Press, and Political Accountability Princeton: Princeton University Press.Bawn, Kathleen and Frances Rosenbluth. 2003. “Coalition Parties versus Coalitions of Parties: How Electoral Agency Shapes the Political Logic of Costs and Benefits.” Paper presented at the 2003 Annual Meeting of the American Political Science Association. Bawn, Kathleen and Francis Rosenbluth, 2006. “Short versus Long Coalitions: Electoral Accountability and the Size of the Public Sector,” American Journal of Political Science. Broder, David S. 1972. The Party's Over: The Failure of Politics in America New York: Harper & Row. Cohen, Marty 2005. Moral Victories: Cultural Conservatism and the Creation of a New Republican Congressional Majority Doctoral Dissertation. UCLA. Cohen, Marty, David Karol, Hans Noel and John Zaller. 2001. “Beating Reform: The Resurgence of Party in Presidential Nominations.” Paper prepared for the 2001 American Political Science Convention, September 2001, San Francisco, Calif. Cohen, Marty, Hans Noel and John R. Zaller 2004. “Without a Watchdog: The Effect of Local News on Political Polarization in Congress.” Paper presented at the Annual Meeting of the American Political Science Association. Chicago Cox, Gary and Mathew McCubbins 1993. Legislative Leviathan. Berkeley: University of California Press. Cox, Gary W. and Keith T. Poole. 2002. “On Measuring Partisanship in Roll-Call Voting: The U.S. House of Representatives, 1877-1999.” American Journal of Political Science. 46 477-489. Gienapp, William E. 1988. The Origins of the Republican Party, 1852-1856 Oxford: Oxford University Press. Huber, Gregory A. and Kevin Arceneaux 2006. “Uncovering the Persuasive Effects of Presidential Advertising” 40 Johnston, Richard, Michael G. Hagen and Kathleen Hall Jamieson 2004. The 2000 Presidential Election and the Foundations of Party Politics Cambridge: Cambridge University Press. Karol, David 2005. Coalition Management: Explaining Party Position Change in American Politics Doctoral Dissertation. UCLA. Key, Jr., V. O. 1952. Politics, Parties, and Pressure Groups New York: Thomas Y. Crowell Company. Key, V.O. 1966. The Responsible Electorate: Rationality in Presidential Voting 1936–1960 Cambridge: Harvard University Press. Koger, Gregory, Seth Masket and Hans Noel 2006. “Partisan Webs: Using the Address Market to Study Extended Party Networks” Paper prepared for the Annual meeting of the Midwest Political Science Association. Masket, Seth 2004. A Party by Other Means: The Rise of Informal Party Organizations in California Department of Political Science. Doctoral Dissertation. UCLA May, John D. 1965. “Democracy, Organization, Michels.” The American Political Science Review. 59 417-429 Mayhew, David R. 1974. Congress: The Electoral Connection New Haven, London: Yale University Press. Mayhew, David R. 1986. Placing parties in American politics: organization, electoral settings, and government activity in the twentieth century Princeton: Princeton University Press. Noel, Hans 2005. “Parties, Ideology and the Emergence of the Anti-Slavery Coalition.” Paper prepared for the annual meeting of the American Political Science Association Polsby, Nelson W. and Aaron B. Wildavsky 1968. Presidental Elections: Strategies of American Electoral Politics New York: Charles Scribner's Sons. Pomper, Gerald 1966. Nominating the President; the politics of convention choice, with a new postscript on 1964. Evanston: Northwestern University Press. Rosenstone, Steven 1983. Forecasting Presidential Elections New Haven: Yale University Press. Schattschneider, E.E. 1942. Party Government Westport, CT: Greenwood Press. Schwartz, Thomas 1989. Why Parties? UCLA manuscript. Snyder, James M. Jr., and Tim Groseclose. 2000. “Estimating Party Influence on Congressional Roll-Call Voting.” American Journal of Political Science. Todorov, A., A. N. Mandisodza, A. Goren and C. C. Hall. 2005. “Inferences of competence from faces predict election outcomes.” Science. 308 1623-1626. Weber, Max 1946. Science as Vocation: From Max Weber: Essay in Sociology. New York: Oxford University Press. 41 Weingast, Barry. 1979. “A Rational Choice Perspective on Congressional Norms.” American Journal of Political Science. 23 Zaller, John.2004. “Floating Voters in U.S. Presidential Elections, 1948-2000.” Studies in Public Opinion : Attitudes, Nonattitudes, Measurement Error, and Change. P. Sniderman and W. Saris, eds. Princeton: Princeton University Press. 42