PROMISSORY REPRESENTATION

Promissory representation, the traditional model, follows the classic principal–agent format. The problem for the principal (in Bristol or Ohio) is one of keeping some control over the agent (in London or Washington). The problem in politics does not differ greatly from the problem of keeping any economic agent responsive to the desires of the principal. Economic history and theory have focused recently on the problem of long-distance trade when there was no governmental infrastructure to enforce contractual arrangements. In the Mediterranean in the fourteenth century, this situation necessitated either kinship ties or above-market payment rates to ensure that ships loaded with the surplus value of thousands of workers actually returned with the goods received in trade (see Greif 1993). When control (as in a seabound ship) or information (as in relations with an expert) is asymmetric, the problem for the principal is to make sure that the agent (the captain, the lawyer, the accountant) acts to further the interests of the principal (the merchant, the client). So too in political representation, both descriptive and normative writers have perceived the problem as one of the voters in a district keeping legal or moral control over their distant representatives. The normative understanding of accountability in promissory representation is that the representative is “responsible to,” “answerable to,” “bound,” and even “bound by” those voters.²

In the “mandate” version of the model, the representative promises to follow the constituents' instructions or expressed desires; in the “trustee” version the representative promises to further the constituency's long-run interests and the interests of the nation as a whole.

In promissory representation, the power relation from voter to representative, principal to agent, runs forward in linear fashion. By exacting a promise, the voter at Time 1 (the election) exercises power, or tries to exercise power, over the representative at Time 2 (the governing period):

Promissory representation thus uses the standard forward-looking concept of power, as in Robert Dahl's (1957) intuitive “A has power over B to the extent that he can get B to do something that B would not otherwise do” (202–203). Indeed, any definition of power derived, like Dahl's, from Weber ([1922] 1978, 53) will imply this kind of forward-looking intentionality. Dahl's “get” implies both that A acts with intention and that B's action will occur in the future. The power relation follows the simplest version of a principal-agent model, with the voter as principal, statically conceived, trying to exercise power over the representative as agent.

³

Promissory representation works normatively through the explicit and implicit promises that the elected representative makes to the electorate. It works prudentially through the sanction the voter exercises at the next election (Time 3). That sanction is a reward or punishment for acting or failing to act according to the promise made at the previous election (Time 1). Both normatively and prudentially, the electoral audit at Time 3 focuses on whether or not the promises at Time 1 were kept. George Bush thus angered his supporters deeply by breaking an explicit campaign promise (“Read my lips: No new taxes”).⁴

Promissory representation has the advantage that, at least in its more mandated versions, it reflects in a relatively unmediated manner the will (although not necessarily the considered will) of the citizenry. It comes closer than any other model to an ideal in which the simple imprint of the voter's will is transmitted through institutions to an equal exertion of power on the final policy. Although promissory representation has never described actual representation fully, it has been and remains today one of the most important ways in which citizens influence political outcomes through their representatives.

Promissory representation thus focuses on the normative duty to keep promises made in the authorizing election (Time 1), uses a conception of the voter's power over the representative that assumes forward-looking intentionality, embodies a relatively unmediated version of the constituent's will, and results in accountability through sanction.

How we conceive of representation begins to change, however, when we consider the implications of instituting a sanction at Time 3.

ANTICIPATORY REPRESENTATION

For more than a generation now, empirical political scientists have recognized the significance in the representative system of “retrospective voting,” in which the voter looks back to the past behavior of a representative in deciding how to vote in the next election. Yet the normative implications of this way of looking at representation have not been fully explored. Returning to the model of promissory representation, it seems obvious that the power exercised in that model works through the voter's potential sanction of voting a representative out of office at Time 3. This is “retrospective voting.” From the representative's perspective, however, retrospective voting does more than provide the potential retribution for broken promises. It also generates what I call “anticipatory” representation, in which the representative tries to please future voters. Whereas in promissory representation the representative at Time 2 (the period in office) represents the voter at Time 1 (the authorizing election), in anticipatory representation the representative at Time 2 represents the voter at Time 3, the next election. ⁵

In anticipatory representation, what appears to the representative to be a “power relation” thus works not forward, but “backward,” through anticipated reactions, from the voter at Time 3 to the representative at Time 2:

Strictly speaking, the beliefs of the representative at Time 2 about the future preferences of the voter at Time 3, not the actual preferences of the voter at Time 3, are the cause of the representative's actions at Time 2. A later event cannot cause an earlier event. Indeed, the representative's beliefs may turn out to be mistaken. Nevertheless, from the perspective of the representative, the entity that exerts the sanction and thus the control appears to be the voter at Time 3.

The model of anticipatory representation thus requires a concept of power different from traditional, forward-looking, intention-based concepts such as Dahl's or Weber's. It requires a concept of power that can include “anticipated reactions.” We find early formulations of this idea in the writings of Carl Friedrich (1937, 16–17, 1958, 1963, ch. 11), Peter Bachrach and Morton Baratz (1963), and Stephen Lukes (1974). The best formulation for the purposes of this analysis comes from Jack Nagel (1975, 29), who defined power, at the highest level of generality, as a “causal relation between the preferences of an actor regarding an outcome and the outcome itself.” The neutrality of this definition in regard to intention and time make it compatible with anticipatory representation. Unlike Dahl's definition, Nagel's definition allows the anticipated preferences of the voter at Time 3 (that is, the representative's beliefs about those preferences) to cause the actions of the representative at Time 2.

Anticipatory representation directs empirical attention away from the relation between Time 1 (the authorizing election) and Time 2 (the representative's period of service), and toward the relations that arise between the beginning of Time 2 (the representative's period of service) and Time 3 (the next election). When preferences are stable over time, there is no important difference between the voter at Time 1 and Time 3 (Miller and Stokes 1963, 50; Nagel 1975, 24ff). But when preferences are unstable or emergent, the representative has incentives to search during Time 2 for the characteristics of the voter at Time 3. Because this anticipation usually poses an extremely difficult information problem (Stimson, Mackuen, and Erikson 1995, 545), the search prompts attention to public opinion polls, focus groups, and gossip about the “mood of the nation” (Kingdon 1984, 153; Stimson, Mackuen, and Erikson 1995, 544). It also prompts attempts to change the voter at Time 3 so that the voter will be more likely to approve of the representative's actions.

This temporal shift has three implications for empirical description and analysis. First, the model becomes more deliberative. The space between Time 1 and Time 3 becomes filled with reciprocal attempts at the exercise of power and communication, much of it instigated by the representative:

Second, anticipatory representation prompts attention to underlying interests as well as present preferences. Benjamin Page (1978, 221–22), for example, points out that a theory of democracy based on the representative's anticipation of reward and punishment “orients government responsiveness toward fundamental needs and values of the people rather than toward ephemeral or weakly held policy preferences.” Douglas Arnold writes that the representative is better off thinking of the voters in the next election as having “outcome” preferences rather than “policy” preferences (1990, 17, 1993, 409). James Stimson (1995, 545) and his colleagues similarly argue that the information problem involved in rational anticipation encourages representatives to aim at general rather than specific knowledge. If we add to these formulations the idea that voters can change their preferences after thinking about them, we can find a place in empirical theory for the concept of “interests” (defined as enlightened preferences) in what would otherwise be a purely preference-oriented model of political behavior.⁶

Third, following from the first two points, anticipatory representation encourages us to think of voters at Time 3 as educable (or manipulable). Between Time 1 and Time 3 the voters can be “educated” not only by the representative, who seeks and prepares “explanations” of his votes (Fenno 1978; Kingdon 1981), but also—critical for the practice of democracy—by parties, interest groups, media, opposition candidates, and other citizens (Arnold 1990, 1993, 409; Kuklinski and Segura 1995, 15–16; Young 2001). (In the following diagram, groups, media, opposition and other citizens are all demarcated as “G” for “Groups.” The arrow indicates both power and communication.)

Arnold (1993), Stimson et al. (1995), and others have drawn the attention of empirical political scientists to this form of representation. They have done so, however, without emphasizing its deliberative side. Arnold's “alternative control model,” which otherwise describes well the process I call “anticipatory representation,” does not fully capture the crucial elements of continuing communication and potentially changing voter preferences. Arnold (1993, 410) describes citizens in the model statically, as acting “more like spectators who register their approval or disapproval at the end of a performance.” Yet Arnold (1993) himself recognizes that anticipatory representation can be intensely interactive with citizens when he notes that legislators “learn from interest groups, committee hearings, staff members, and other legislators about the policy consequences and the political consequences of specific decisions” (412). Interest groups and committee hearings are both institutions by which citizens communicate their evolving interests and opinions (although not without intervening biases introduced by the selection and medium of communication).

Arnold also describes legislators statically, as “controlled agents.” Although he is right in saying the legislators are not “instructed delegates,” his phrase “controlled agents” does not capture the legislators' role as potential initiators and educators. In contrast, the model of anticipatory representation is in most instances interactive and more continually reflexive. Anticipatory representation derives from a marketplace model, which Arnold (1993, 412) himself adopts when he writes that “movie makers, auto makers, and real estate developers attempt to anticipate and satisfy consumers' preferences.” In the marketplace, customers are not mere “spectators”; nor are entrepreneurs “controlled agents.” Rather, customers actively (if not intentionally) exert power and influence on the marketplace, and entrepreneurs too are active, in searching out and sometimes even creating preferences. Like the customer/entrepreneur relation in the marketplace, the voter/representative relation in anticipatory representation is best conceived as one of reciprocal power and continuing mutual influence.

The temporal shift produced by anticipatory representation has parallel implications for normative theory. Most prominently, it undermines the traditional understanding of accountability. It therefore demands new normative criteria in its place.

The traditional concept of accountability, focusing on the relationship between Time 1 and Time 2, asks whether the representative is doing what the statically conceived constituent wanted the representative to do at Time 1. By substituting the voter at Time 3 for the voter at Time 1, anticipatory representation makes the voter at Time 1 irrelevant. If we think of the representative as an entrepreneur, anticipating future customers' preferences, the forces that make the representative “accountable” are all forward looking. Yet it would seem strange to say that the representative was accountable to the voter at Time 3.

The argument that anticipatory representation undermines traditional notions of accountability will seem counterintuitive, because, of all the models I introduce here, anticipatory representation is most intimately related to those traditional notions. The desire for reelection is usually, and quite reasonably, interpreted as simply a mechanism for insuring the fidelity of the representative to the voter's wishes, making no distinction between the voter at Time 1 and the voter at Time 3. Indeed, if the voter at Time 3 does not differ from the voter at Time 1, then we can think of the voter at Time 3 as simply doling out the reward or punishment to enforce the power relation in promissory representation.

Most theorists and most members of the public still envision representation through the traditional model of promissory representation, in which the voter's power works forward and the representative's attention looks backward. The public's advocacy of term limits, for example, adopts this static feature of the traditional model. The voters fear that the farther away the representative gets from home, literally and figuratively, the weaker the tether that holds that representative to them. The voters want their “hooks” in the representative to be strong. In the intensity of that desire, they seem willing to forgo the reelection incentive. Their implicit calculus seems not to include the incentives built into Time 3.

But the shift in temporal emphasis in anticipatory representation brings unexpected normative changes in its wake. To the degree that we think of the legislator as representing the voter at Time 3, we turn the legislator into a Shumpeterian entrepreneur, motivated to try to attract the votes of future customers. As we have seen, in this conception, strictly speaking, the traditional principal–agent model disappears. We do not think of an economic entrepreneur as an agent, with the future customers as principals. A representative trying to anticipate the desires of voters at Time 3 has a prudential, not a moral, relationship to those voters. To the degree that the representative wants to be reelected, he or she will see pleasing the voters (and funders) at Time 3 as the means to that end. Whereas in traditional accountability, we would say that the representative “ought” to do what he or she had promised the voters at Time 1, we do not say that the representative “ought” to try to please the voters at Time 3. In this respect, purely prudential incentives have replaced a combined moral and prudential imperative.

Replacing morality with prudence in the incentive structure of anticipatory representation leads us to judge the process with new normative criteria. It makes us shift our normative focus from the individual to the system, from aggregative democracy to deliberative democracy, from preferences to interests, from the way the legislator votes to the way the legislator communicates, and from the quality of promise-keeping to the quality of mutual education between legislator and constituents.

Anticipatory representation forces normative theory to become systemic. In most anticipatory representation, the better the communication between voter and representative in the interval between Time 1 and Time 3, the better the representation. A representative could in theory accurately anticipate the desires of the voter at Time 3 without any mutual communication. In practice, representatives usually initiate and welcome the opportunity to communicate with voters, both to anticipate their preferences at Time 3 and to influence them. The quality of that mutual communication then depends only in small part on the dyadic efforts of the representative and the constituent. It depends much more on the functioning of the entire representative process—including political parties, political challengers, the media, interest groups, hearings, opinion surveys, and all other processes of communication. Each of these has important functions in an overall process of what might be called “continuing representation.” Normative theory should ask, and empirical political science should try to answer, how well the entire representative system contributes to ongoing factually accurate and mutually educative communication (see Williams 1998 and Young 2000, 128, 130 on interaction; Thompson 1988 on representation over time).

Focusing on the changes in voter and representative between Time 1 and Time 3 also underlines the deliberative function of representation. Recognizing that the representative's initiatives have the potential to change as well as to anticipate voters at Time 3, normative theorists should be able to help empirical political scientists ask whether those changes are best described as “education” or “manipulation.⁷

Manipulation may be distinguished by the intent to deceive or create conditions of choice leading others to make a choice not in their interests (see Lukes 1974). Beyond nonmanipulation, the quality of education can be judged by the deliberative criteria of whether the mutual interaction between Time 1 and Time 3 makes the voters at Time 3 (1) more or less aware of their underlying interests and the policy implications of those interests and (2) more or less able to transform themselves in ways that they will later consider good (including, when appropriate, becoming more concerned with the common interest).

Education, in short, is a form of what I will call “influence” and manipulation a form of what I will call “coercive power.” Within Nagel's broad understanding of power as preferences causing outcomes, we may distinguish analytically between these two forms. Influence, marked by (relatively) common interests on the issue between influencer and influenced, is exercised through arguments on the merits. Coercive power, marked (except in paternalism) by a conflict of interest between power exerciser and recipient, has two subtypes: “The threat of sanction,” which involves the will of the actor subject to power, and “force,” which includes not only physical force but any structuring of alternatives that constrains the choices of the actor subject to power regardless of that actor's will. “Education” may be conceived as a form of influence, as it works through arguments on the merits and is by definition in the recipients' interests. “Manipulation” may be conceived as a form of force, as it occurs, by definition, against the recipients' interests without their recognizing characteristics of the situation that might have led them to take another action.⁸

None of these forms of power is easy to operationalize, because their definitions involve contests over what is and what is not in an individual's interests.

Normative theorists are currently working to define the appropriate standards for the use of coercive power and influence. Regarding coercive power, the normative theory appropriate for aggregative models of democracy mandates that each voter's preferences should have roughly equal coercive power over the outcome. In deliberation, in contrast, the ideal is the absence of coercive power.⁹

In deliberation, influence can legitimately be highly unequal (at least under conditions in which the unequal exercise of influence does not undermine a rough equality of respect among participants, foreclose further opportunities to exercise equal power, or deny any of the participants the opportunity to grow through participation). Knight and Johnson (1998) argue convincingly for an ideal of “equal opportunity of access to political influence” in democratic deliberation. But even that ideal is a default position, holding unless good reasons can be given for unequal access to influence. In formal representation, for example, citizens for good reasons place the representative in a position of greater potential influence and coercive power than most constituents. When a representative uses that greater coercive power in a deliberation, e.g., to set the agenda, that act is not automatically normatively wrong (as suggested by both ideals of equal access to influence and absence of coercive power) but should be judged by the three criteria, appropriate to deliberation, of nonmanipulation, illuminating interests, and facilitating retrospectively approvable transformation.

Unfortunately for analyses that try to be purely “objective,” questions regarding voters' interests, in contrast to their preferences, are not susceptible to certain resolution. They are “essentially contested” (Gallie 1962). They are nevertheless the right questions to ask. These questions force the observer to consider whether the process of mutual communication with the representative deepens the base on which the voters' preferences rest, or instead introduces misleading considerations or emphases that, given adequate information and the time for adequate reflection, the voters would reject.

At the moment, the existing representative apparatus in the United States does not facilitate well the processes of mutual education, communication, and influence. For example, when William Bianco (1994, 51) asked members of Congress whether they thought they could explain to their constituents a vote (against the repeal of Catastrophic Coverage for health insurance) that they considered a vote for good public policy, many found that their attempts at education only made their constituents angry.¹⁰

In this case, some constituents (whose private policies covered much of what the bill would provide) had far greater access to influence than others. Some political entrepreneurs deceived the public, probably intentionally (King and Scott 1995). Critically, representatives had neither the political space nor the time to explain their reasoning to their constituents and be educated in turn. The citizens did not have forums in which they could discuss together all aspects of the matter. The deliberative process thus fell far short of meeting not only the criteria of equal opportunity for access to influence and nonmanipulation but also the criteria of interest clarification and (less relevantly here) retrospectively approvable transformation, which might have justified unequal access.

In the case of Catastrophic, political parties, the media, and the relevant interest groups played only minor roles in rectifying distortions in the process of representation. Yet in a polity the size of the United States, these intermediaries play a crucial role in the larger system of representation. By emphasizing the distance between the representative and the voter, the traditional model of promissory representation puts little weight on the quality of communication between the two. In contrast, the incentive structure behind anticipatory representation has created an entire apparatus of opinion polling, focus group, and interest group activity that deserves closer normative scrutiny. Rather than treating opinion polls and focus groups as tools of manipulation and interest groups as no more than the tool of “special interests,” an empirical analysis driven by appropriate normative concerns should ask how well these institutions, along with opposition candidates, political parties, and the media, avoid the biases of unequally funded organizational forms and how well they serve the normatively worthy purposes of mutual communication and education.¹¹

Such a focus would inevitably draw one away from the dyadic representative–constituent relation and toward the larger system of multi-actor, continuing representation.

In short, if in anticipatory representation the representative simply anticipated the preferences of the voter at Time 3 and made no move to change those preferences, the aggregative norms of equal power per voter that underlie the promissory model would need no supplementation. But if, as seems to be the case in almost all actual instances, representatives use their power and influence to affect the preferences of voters at Time 3, the norms of good deliberation must come into play, and we must ask whether the criteria of nonmanipulation, interest clarification and retrospectively approvable transformation that justify unequal access to influence are being met or at least approached.

Anticipatory representation thus focuses on the prudential incentive to please the voter in the next election (Time 3), uses a conception of the voter's power over the representative that allows anticipated reactions, replaces the constituent's transmission of will with the representative's desire to please, and shifts normative scrutiny from the process of accountability to the quality of deliberation throughout the representative's term in office.

GYROSCOPIC REPRESENTATION

I have given the label “gyroscopic representation” to a conception of representation that not only differs from, but is to some degree incompatible with, anticipatory representation. Others have called this representation by “recruitment” (Kingdon 1981, 45), by “initial selection” (Bernstein 1989), or by “electoral replacement” (Stimson et al. 1995).¹²

In this model of representation, voters select representatives who can be expected to act in ways the voter approves without external incentives. The representatives act like gyroscopes, rotating on their own axes, maintaining a certain direction, pursuing certain built-in (although not fully immutable) goals. As in the other new models of representation introduced here, these representatives are not accountable to their electors in the traditional sense. In this case, the representatives act only for “internal” reasons. Their accountability is only to their own beliefs and principles.

This model can take several forms. In all forms the representative looks within, for guidance in taking action, to a contextually derived understanding of interests, interpretive schemes (“common sense”), conscience, and principles. In the United States, a voter may select the narrowest version of the type, dedicated to a single issue such as the legalization of abortion. Or a voter may select the broadest version, a person of integrity with a commitment to the public good. In general, people often try to select what Fearon (1999, 68) calls a “good type,” with the characteristics of (1) having similar policy preferences to the voter, (2) being honest and principled, and (3) being sufficiently skilled. They explain their choices, for example, with the phrase, “He's a good man” or “She's a good woman” (Fenno 1978, 55; Miller and Stokes 1963, 54).

Character, including adherence to principle, is an important feature on which voters select. But it is not the only feature. In the United States, voters also use descriptive characteristics, along with party identification and indicators of character, as cues by which to predict the representative's future behavior (Popkin 1994). Legislators themselves often adopt this understanding of representation, seeing themselves as having an attitudinal identity with a majority of their constituents (Bianco 1994, 39; Fenno 1978, 115; Kingdon 1981, 45–47). Thus the two principal features that Fearon (1999) enunciates, of having policy preferences similar to the constituent's and being honest and principled, are analytically separable but entwined in practice, because similar policy preferences will not suffice if the representative can be bribed.¹³

In the “party discipline” models characteristic of much of Europe, representatives look within to a set of principles and commitments that derive partly from their own ideals and partly from their commitment to the collective decisions of the party. The representative is also subject to party sanctions for not obeying the party, and the party in turn is subject to sanctions from the voters. I focus here only on the model of gyroscopic representation that prevails in the United States.

In all versions of gyroscopic representation, the voters affect political outcomes not by affecting the behavior of the representative (“inducing preferences,” as in promissory or anticipatory representation), but by selecting and placing in the political system representatives whose behavior is to some degree predictable in advance based on their observable characteristics. Whereas in promissory and anticipatory representation the representative's preferences are induced, in this model the representative's preferences are internally determined. Whereas in promissory and anticipatory representation the voters (at Time 1 or Time 3) cause changes in the representative's behavior, in gyroscopic representation the voters cause outcome changes first in the legislature and more distantly in the larger polity not by changing the direction of the representative's behavior but by placing in the legislature and larger polity (the “system”) the active, powerful element constituted by this representative. The voters thus have power not over the representative, but over the system:

In this form of representation, the representative does not have to conceive of him or herself, in Pitkin's ([1967] 1972) terms, as “acting for” the constituent, at either Time 1 or Time 3. The motivations of the representative can remain a black box. The voter selects the representative based on predictions of the representative's future behavior derived from past behavior and other cues. We may envision the candidates vying for election as a set of self-propelled and self-directed thinking, feeling and acting machines, from which the voter selects one to place in the system. After the selection, the self-propelled machine need have no subsequent relation to the voter. The key to the voter–representative relationship in this model is thus not traditional accountability but deep predictability, in the sense of predicting an inner constellation of values that is, in important respects, like the constituent's own. In some electoral systems, the political party is often far more predictable and easier for voters to relate to their own interests than are individual politicians. In the United States, a politician's personal reputation, descriptive characteristics, and character (as the voters judge it) provide deep predictability above and beyond the predictor of party identification.

In the United States, gyroscopic representation forms a relatively large part of the representative process. As John Kingdon (1981, 45) writes, “The simplest mechanism through which constituents can influence a congressman is to select a person initially for the office who agrees with their attitudes.” Approximately three-quarters of the time Kingdon (1981, 45) found no conflict between what a majority of the constituency wanted and the personal attitudes of their member of Congress. Gyroscopic representation (or representation by recruitment) could therefore comprise as much as three-quarters of the dynamic of representation in the United States Congress. Robert Bernstein (1989) agrees with this assessment, dubbing the prevailing fixation on what I call promissory representation and anticipatory representation “the myth of constituency control.” In the most elegant analysis to date, Stimson et al. (1995) provide data suggesting that in the United States Senate and presidency, gyroscopic representation (their “electoral replacement”) is the most important mechanism by which the representatives respond to public opinion changes. In the House of Representatives, their data suggest, the most important mechanism is anticipatory representation (their “rational anticipation”).

Like anticipatory representation, gyroscopic representation has some ties to the traditional form of accountability postulated in promissory representation, but there are also crucial differences. In gyroscopic representation, the representatives do have a normative responsibility to their constituents not to lie about the characteristics on which they are being selected at election time. But in the gyroscopic model the deeper accountability of the representatives is to themselves or (particularly in electoral systems outside the United States) to the political party with which they identify. They are not expected to relate to their constituents as agents to principals. As Kingdon (1981, 46) puts it, in this model the member of congress “never even takes [the constituency] into account.” Or, as Fearon (1999, 56) writes, “electoral accountability is not necessary.” The fiduciary component to the relation is weak. The tether to the voter at Time 1 is almost nonexistent.

Gyroscopic representation also differs from Burke's “trustee” form of representation. Burke ([1774] 1889) envisioned the representative as a statesman, concerned with interests rather than mere preferences and with the interests of the entire nation rather than the district.¹⁴

Yet in gyroscopic representation, the voter may select a representative only because both voter and representative share some overriding self-interested goal, such as lowering taxes. Or the voter may select a representative with many of the voter's own background characteristics, on the grounds that such a representative will act much the way the voter would if placed in the legislature. The point for the voter is only to place in the system a representative whose self-propelled actions the voter can expect to further the voter's own interests. Burke's “trustee” conception thus comprises one subset within the larger concept of gyroscopic representation.

The gyroscopic model does resemble Burke's trustee conception in one important respect. Having decided that the representative already wants, for internal reasons, to pursue much the same course as the one the voter wants, the voter often expects the representative (or the party) to act with considerable discretion in the legislature. This expectation opens the door to creative deliberation and negotiation at the legislative level. Compromises, changes of heart, and even the recasting of fundamental interests are all normatively permitted.

As we have seen, traditional accountability is irrelevant in the gyroscopic model. In the pure form of the model, as Kingdon points out, the representative never takes the constituency into account and is not expected to do so. The quality of ongoing communication between representative and constituent is also irrelevant. In the pure form of the model, as Kingdon also points out, the ongoing communication between the representative and the constituent can, even ideally, be nil. The normative process of judging this form of representation thus requires criteria that differ from those of traditional accountability.

One critical criterion, deliberation at authorization, requires normatively estimating the quality of deliberation among constituents and representatives before and at Time 1, the authorizing election. Good deliberation at this moment would result in voters achieving both developed understandings of their own interests and accurate predictions of their chosen representatives' future behaviors. Good deliberation requires that representatives not intentionally deceive the public as to their future behavior. The voter's aim is to discern and select on the criterion of commonality of interests between the representative and the constituent (see Bianco 1996).

A second criterion, ease of maintenance and removal, requires that the voters be able at periodic intervals to reenter the system, either perpetuating its current direction by maintaining their self-propelled representatives in office or changing that direction by removing one representative and inserting another. Term limits, which make sense in a model of promissory representation, make little sense either for anticipatory representation or for gyroscopic representation. Term limits make it impossible to maintain one's chosen representative in the system.

In short, the normative criteria appropriate for gyroscopic representation are good systemwide deliberation at the time of selection (the authorizing election) and relative ease in maintaining one's selected representative in office or removing that representative and placing another in the system. Gyroscopic representation stresses the representative's own principles and beliefs, sees the voter as having power not over the representative but over the system (by inserting the representative in that system), and shifts normative scrutiny from traditional accountability to the quality of deliberation in the authorizing election.

SURROGATE REPRESENTATION

Surrogate representation is representation by a representative with whom one has no electoral relationship—that is, a representative in another district. As with the other forms of representation, I am not the first to notice the importance of this kind of representation in the United States today. Robert Weissberg described it in 1978 as “collective representation,” and John Jackson and David King in 1989 called something similar “institutional” representation. Edmund Burke had a version he called “virtual” representation, but Burke's concept focused on morally right answers, wisdom rather than will, relatively fixed and objective interests, and the good of the whole, which is only one of many possible goals for surrogate representation.¹⁵

In the United States today, individuals and interest groups representing individuals often turn to surrogate representatives to help advance their substantive interests, including their ideal-regarding interests. A member of Congress from Minnesota, for example, may lead the Congressional opposition to a war opposed by significant numbers of voters in Missouri and Ohio whose own representatives support the war. The situation has changed from the time when territorial representation captured many of a voter's most significant interests, but in the United States the representational system has not changed with it. In the United States, surrogate representation—a noninstitutional, informal, and chance arrangement—is the preeminent form of nonterritorial representation.

For the affluent (or the organized, e.g., through labor unions), surrogate representation is greatly enhanced by the possibility of contributing to the campaigns of representatives from other districts. Individual candidates, political parties, and many other political organizations as a matter of course solicit funds from outside their districts. Citizens with ample discretionary income find many of their most meaningful instances of legislative representation through what one might call “monetary surrogacy.”

Surrogate representation, both state- and nationwide, plays the normatively critical role of providing representation to voters who lose in their own district. Because both federal and state electoral systems use single member districts, with first-past-the-post, winner-take-all majority elections, citizens whose preferred policies attract a minority of voters in their own districts could theoretically end up with no representation at all in the legislature. Yet with sufficient geographic clustering, the interests and perspectives that lose in one district will win in another, so that voters in the minority in District A will have surrogate representation through the representative of District B. In electoral systems structured this way, the accidental supplement to existing institutions provided by surrogate representation is crucial to democratic legitimacy. As we shall see, if serendipity did not produce enough surrogate representation to meet systemic criteria for legitimacy, the electoral system as a whole would not withstand normative scrutiny.

In the kind of surrogate representation that is not anchored in money or other contributions (“pure” surrogate representation), there is no relation of accountability between the representative and the surrogate constituent. Nor is there a power relation between surrogate constituent and representative:

The only power relation (in the sense of the threat of sanction or the use of force) arises between those who contribute money or other goods and the representatives to whose campaigns they contribute. In a relation of monetary or contributing surrogacy, the contributor exerts power through exacting promises as in traditional representation, through anticipated reactions as in anticipatory representation, and through placing in the system a legislator who will predictably act in certain ways as in gyroscopic representation. Because all the power that is exercised in any surrogate representation works through monetary or other contributions and through contributors rather than voters, surrogate representation in the United States today embodies far more political inequality than does even the traditional legislator–constituent relation.

Yet even without the fear of losing monetary or other contributions, and without any formal accountability, surrogate representatives sometimes feel responsible to their surrogate constituents in other districts. Legislators deeply allied with a particular ideological perspective often feel a responsibility to nondistrict constituents from that perspective or group.

That sense of surrogate responsibility becomes stronger when the surrogate representative shares experiences with surrogate constituents in a way that a majority of the legislature does not. Representatives who are female, African American, or of Polish ancestry, who have a child with a disability, or who have grown up on a farm, in a mining community, or in a working-class neighborhood, often feel not only a particular sensitivity to issues relating to these experiences but also a particular responsibility for representing the interests and perspectives of these groups, even when members of these groups do not constitute a large fraction of their constituents. Feelings of responsibility for constituents outside one's district grow even stronger when the legislature includes few, or disproportionately few, representatives of the group in question.¹⁶

Representative Barney Frank, a Democrat from Massachusetts, consciously sees himself as a surrogate representative for gay and lesbian citizens throughout the nation. Frank, who is himself openly gay, has a sympathetic district constituency: “My constituents at home understand my position. Issues concerning gay and lesbian discrimination are important to me.” He points out that he is able to play this role because it does not take a great deal of time and therefore does not detract much from what he does for his district. Frank takes his surrogate responsibilities seriously. He believes that his surrogate constituents nationwide “know I understand their concerns…. I have a staff with three openly gay, talented lawyers who feel committed to helping this problem at large.”¹⁷

He receives mail from gay and lesbian citizens across the nation “regarding their concerns about gay rights and discrimination,” and he feels a special responsibility to that group, because he is one of the few openly gay members of Congress. In his case, this sense of responsibility is increased because the constituents who write him from around the nation are often not in a position, due to prejudice against them, to become politically active on their own.¹⁸

The relation of a surrogate representative with surrogate constituents can also be somewhat deliberative. In addition to their contributions of money, in-kind services, and volunteer time, which foster a form of power relation, groups represented in surrogate fashion may provide information and expertise. (The moral approbation or disapproval that they also provide may be conceived in part as information and in part as an exercise of power.) Surrogate representatives may consult with group members, particularly those who have some formal or informal claim to represent others of the group, so that information and insights flow both ways.

Although dyadic, district-based, representative–constituent accountability is completely absent in pure surrogate representation, we can nevertheless develop normative criteria to judge the degree to which, on a systemic basis, that surrogate representation meets democratic standards. The most obvious criterion is that the legislature as a whole should represent the interests and perspectives of the citizenry roughly in proportion to their numbers in the population. But to this larger criterion we must enter certain caveats.

First, the aggregative aims of democracy require that the most conflictual interests be those on which most effort is made to achieve proportionality in representation. When interests conflict in ways that cannot be reconciled by deliberation, the Anglo-American theory of democracy that has evolved since the seventeenth century rests the fairness of the conflict-resolving procedure on some approximation to equal coercive power among the parties. The norm of “one person/one vote” implies the equal individual power of a vote in a direct democracy and equal proportional power in a representative democracy. The more important the conflict, the more vital becomes a proportional representation of the relevant interests.

Second, the deliberative aims of democracy require that the perspectives most relevant to a decision be represented in key decisions. Such perspectives do not necessarily need to be presented by a number of legislators proportional to the number of citizens who hold those perspectives.¹⁹

The goal is to produce the best insights and the most relevant information, through mutual influence, which in deliberation may legitimately be unequal, not through coercive power, which ideally should be absent.

Deliberative goals may also justify some of the inequality currently characteristic of surrogate and other forms of representation. When the deliberative mechanisms built into an electoral system work well, they should select, through “the force of the better argument,” against, at the very least, the least informed political positions in the polity. Accordingly, the representatives in the legislature who advocate these positions should be fewer proportionately than the number of citizens who hold that position. Good deliberation should work through the electoral process as well as through other processes of mutual education to winnow out the least informed ideas, leaving the best in active contest.²⁰

The current surrogate selection process in the United States departs significantly from the democratic standard. Although existing electoral systems do to some degree select the best ideas, surrogate systems, even more than direct elections, select primarily for the best financed ideas and interests. In the United States inequalities of this sort are often justified on the grounds that they reflect freedom of “speech,” as conveyed through monetary contribution. But unequal contributions to surrogate representatives are, I would argue, not justified on the grounds of either adversary fairness (providing proportional representation to conflicting interests) or deliberative efficacy (providing some representation for relevant perspectives on a decision).²¹

The normative questions to be asked with regard to surrogate representation differ from the questions posed by traditional accountability. In surrogate representation, legislators represent constituencies that did not elect them. They cannot therefore be accountable in traditional ways. As in gyroscopic representation, the legislators act to promote their surrogate constituencies' perspectives and interests for various reasons internal to their own convictions, consciences, and identities. Or they act to assure the continuous flow of dollars into their campaigns. The normative question for surrogate representation is not, therefore, whether representatives accurately reflect the current opinions or even the underlying interests of the members of their constituencies. Rather, it is whether, in the aggregate, each conflicting interest has proportional adversary representation in a legislative body (Weissberg 1978, esp. 542) and each important perspective has adequate deliberative representation. Such a normative analysis must involve a contest regarding what interests most conflict (and therefore most deserve proportional representation) in aggregation and what perspectives count as important in deliberation.²²

In short, surrogate representation must meet the criteria for proportional representation of interests on relatively conflictual issues (an aggregative criterion) and adequate representation of perspectives on matters of both conflict and more common interest (a deliberative criterion). Surrogate representation thus focuses not on the dyadic relation between representative and constituent but on the systemwide composition of the legislature, sees the represented as exercising no power over either the representative or the system except when the represented makes a (usually monetary) contribution to the representative, and shifts normative scrutiny from constituent-oriented accountability to systemic inequities in representation.

DELIBERATIVE, SYSTEMIC, AND PLURAL NORMATIVE CRITERIA

Table 1 summarizes some of the characteristics of these different forms of representation.²³

When empirical political scientists want to answer the question of how well a political system meets democratic norms, they need a democratic theory that will clarify those norms in ways that make it easier to tell when real-world situations conform to or violate them. In the field of United States legislative studies, the democratic norms regarding representation have often been reduced to one criterion: Does the elected legislator pursue policies that conform to the preferences of voters in the legislator's district? This criterion is singular, aggregatively oriented, and district-based. In contrast, this analysis advocates plural criteria (cf. Achen 1978; Beitz 1989). It further suggests that some of these criteria should be deliberatively-oriented and systemic.

From a deliberative perspective, even promissory representation requires good deliberation to ascertain whether or not representatives have fulfilled their promises or have persuasive reasons for not doing so. Anticipatory representation requires good deliberation between citizens and representatives in the period of communication between elections whenever—as is almost always the case—a representative tries to influence the voter's preferences by the time of the next election. Gyroscopic representation requires good deliberation among citizens and between citizens and their representatives at the time the representative is selected. Surrogate representation requires not only equal gladiatorial representation of the most important conflicting interests in proportion to their numbers in the population but also good deliberative representation of important perspectives.

Each form of representation should also be judged by its contribution to the quality of deliberation in the legislature. In anticipatory representation, a good quality of communication among citizens, groups, and representatives between elections probably improves the quality of deliberation within the legislature. In contrast, one form of gyroscopic representation—based on voters' choosing a representative whom they expect to pursue a vision of the public interest—facilitates good legislative deliberation not by mutual continuing contact and education but by selecting individuals likely to deliberate well and leaving them free to pursue that goal as they think fit. Surrogate representation contributes to good legislative deliberation by making it more likely that varied and important perspectives will be included.

Although a normative judgment on each of these forms of representation involves judging the quality of the deliberation that they produce or that produces them, political theorists are currently only gradually working out what the criteria for good deliberation should be. The standard account is that democratic deliberation should be free, equal, and rational or reasonable. As we have seen in the case of equality, however, each of these characteristics needs greater specification, because not all of the ordinary language meanings of these words ought to apply to the deliberative case. Democratic deliberation should be free in the sense of open to all relevant participants (much hangs here, as elsewhere, on the definition of “relevant”). It should ideally come as close as possible (in a world created by and suffused by power) to a situation in which coercive power has no role and the only “force” is that of the better argument. It should ideally allow equal opportunity of access to influence for all constituents, except where good reasons can be given for unequal opportunity. It should facilitate the expression and processing of relevant emotions as well as cognitions. It should be nonmanipulative. And it should both clarify and appropriately transform individual and collective interests in the directions of both congruence and conflict.²⁴

None of these criteria replace the criterion of constituent-representative congruence. They add to it. Indeed, congruence of a sort is a factor in each of the forms of representation. It is most obvious in promissory representation, where one would expect explicit promises to reflect points of congruence between constituent preferences and a representative's future actions. It applies in anticipatory representation to the reelection, where one would expect constituents to have moved both toward and with the representative's positions and the representative to have moved similarly both toward and with the constituents. In gyroscopic representation one would expect greater congruence to the extent that the representative was elected descriptively to duplicate the median voter but less to the extent that the representative was elected to behave as a principled notable. In surrogate representation, norms of congruence, when applicable, apply to the polity as a whole.

None of the recently identified forms of representation, however, involves accountability in its classic form. In anticipatory representation, strictly interpreted, the representative acts only as entrepreneur, preparing to offer and offering a product to a future buyer. In gyroscopic representation, strictly interpreted, the voter selects a representative who then acts purely autonomously. In pure surrogate representation, there need be no relation at all between the representative and the individual constituent. These three forms of representation supplement the traditional model of promissory representation, which does involve accountability in its classic form. They do not replace the traditional model; nor do they replace the concept of accountability. As legitimate and useful supplementary forms of representation, however, they require separate normative scrutiny.

In most respects, these models of representation are compatible with one another and with promissory representation. They have complementary functions for different contexts and can, thus, be viewed as cumulative, not oppositional. Compatibly, they direct attention to deliberation at different points in the representative system: to the moment of election, between elections, and in the legislature. Compatibly, they all require each voter's interests to have equal weight in contexts of conflicting interests, although promissory representation comes closest to the normative standard of direct democracy, in which the people themselves rule. Compatibly, surrogate representation provides at the national level elements required for systemic democratic legitimacy that the other three forms do not provide. Gyroscopic representation is most appropriate for uncrystalized interests and changing situations but requires considerable constituent trust, which many situations may not warrant. Promissory representation requires little open-ended trust but works badly in situations of rapid change. Anticipatory representation requires little trust and easily accomodates change, but produces incentives for short-term thinking and manipulation focused on the next election.

In a few respects, the models come in conflict. Most importantly, promissory representation restricts the representative's action after election, while gyroscopic representation frees it. Anticipatory representation attracts entrepreneurs; gyroscopic representation, public-spirited notables. Certain functions that might be thought compatible in a division of labor (e.g., gyroscopic representation requiring considerable constituent trust and anticipatory representation relatively little trust) might, from another point of view, be considered conflicts (institutions that assume little trust sometimes drive out institutions that assume greater trust). Other conflicts may become visible over time.

These forms of representation are not mutually exclusive. Moreover, they may interact over time with one another. An anticipatory representative may become a promissory representative at the next election. A legislator may start as a gyroscopic representative and, wings clipped and some trust lost, become a promissory representative. The preferences that constituents express at Time 1 in promissory representation may be the product of earlier anticipatory, gyroscopic, or surrogate processes.²⁵

Although in some respects the normative criteria for judging these forms of representation are additive, the plural criteria of this analysis do not require the models to be fully congruent with one another, any more than the separate normative mandates of freedom and equality need to be congruent. As a consequence, what representatives ought to do when faced with constituent preferences that are not in the constituents' long-term interests or not compatible with the good of the whole is, from the perspective of representational theory, indeterminate. Representatives may legitimately act in several ways, as long as they respect moral norms and the norms appropriate for the model, or combination of models, they are following.