Chapter 1
Coordination and Convention
Edgar ⌠like a true politician, concurred with the prevailing party.
âHume, History, 1.99
Human beings have certain interests in common (we can for now ignore what they are). But since the social and political institutions that we have an interest in supporting are advantageous not individually but collectively, which institutions deserve our support depends on which institutions all other people believe, or can be brought to believe, deserve their support. Many great problems of high politics can thus be seen as problems of coordination. When the status quo or âsocial normâ solution is doubtful or contested, they become problems of authority, since only authority can adjudicate the norm. When the convention of authority itself is doubtful or contested, they are problems of coordinating on authority.
This chapter outlines briefly the problem of coordination and the solution to itâconventionâthat is most relevant in the context of large-scale politics and government. The goal is not to add to a huge literature on coordination and convention but to sketch for non-specialists what these concepts mean and why they matter, so as to explain why we should care about Humeâs profound contribution to their study. Here I endorse Russell Hardinâs claim that Hume has been neglected partly because he was so far ahead of his time: the problems he was addressing were not generally recognized as problems until two centuries later.1 For that very reason, theorists schooled in the modern âcanonâ of great works in political theory but not in recent economics and positive political theory may not be familiar with coordination, or may assume (as indeed formal modelers often encourage us to assume) that coordination need only interest those who think that technical or mathematical models are the best method for treating all political questions.
Coordination theory, that is, divides political theorists into those who assume that no intelligent theorist can fail to know the literature on it and those who are very intelligent but have never seen the point of learning the first thing about it. This chapter will attempt to explain coordination in a way that will be clear and interesting to the second group: those inclined to doubt that mathematics can make sense of politics, or at least wonder whether whatever sense it might make is worth one paragraph of Montesquieu. Conversely, even specialists may benefit from (as well as perhaps disagreeing with) this chapterâs suggestion that a qualitative, narrative account is likely to be more useful than formal models in understanding and explaining the kind of coordination relevant to the fundamentals of political authority. And they may also benefit from considering some unfamiliar political implications of coordination that Hume noted but we often forget. Though coordination problems are often seen as a subset of formal theory, the crucial contribution of Humeâs political theory is to explain, and help solve, coordination problems on scales, and in situations, that the usual tools of formal theory do not address (as well as aspects of those problems, including normative ones, on which formal theory as such is deliberately silent). Though game theoristsâ contributions to coordination theory are of course indispensible, the contribution of Humeâs History begins where those of most contemporary formal theory leave off.
The Problem of Coordination
The modern inventor of coordination theory was Thomas Schelling. I cannot improve on his classic examples. Here is the first one.
Figure 1. Map and following quotation by Thomas Schelling. Redrawn from The Strategy of Conflict by Thomas Schelling, Harvard University Press 1960, reprint 1980. pg. 54.
Two people parachute unexpectedly into the area shown, each with a map and knowing the other has one, but neither knowing where the other has dropped nor able to communicate directly. They must get together quickly to be rescued. Can they study their maps and âcoordinateâ their behavior? Does the map suggest some particular meeting place so unambiguously that each will be confident that the other reads the same suggestion with confidence?
Among the students to whom Schelling posed the question, almost all chose the bridge. In my own more recent trials, almost all still do. The bridge jumps out for its prominence but even more for its uniqueness. A crossroads might be a natural meeting point, but the map has many of those and only one bridge.2 Two more quick examples, among many that Schelling uses:3
You are to meet somebody in New York City. You have not been instructed where to meet; you have no prior understanding with the person on where to meet; and you cannot communicate with each other. You are simply told that you will have to guess where to meet and that he [sic] is being told the same thing and that you will just have to try to make your guesses coincide.4
A majority of Schellingâs initial sample, of students in Connecticut in 1960, chose to meet at the information booth in Grand Central Station. In a variant of the problem that asked people to guess a meeting time as well as a place, almost everyone settled, and still settles, on twelve noon. Another example:
Write some positive number. If you all write the same number, you win.
Most in Schellingâs sample, on reflection, picked the number one; the rest almost all picked powers of ten. One more example: In March 2010, I asked a group of prospective Public Policy students to name any living human being; each student would âwinâ if he or she picked the choice most popular with the rest. My prediction that almost everyone would pick Barack Obama was not difficult and not falsified.
The bridge on the map, Grand Central Station, Barack Obama, and the number 1 are examples of what Schelling called âfocal pointsâ: each provides a âclue for coordinating behavior, some focal point for each personâs expectation of what the other expects him to expect to be expected to do.â While there have been many attempts to formalize focal points in game theory, Schelling resisted them on principled grounds. He argued that the payoffs one writes in a cell will matter for decision-making purposes only to the extent that they are prominent or salient. We might say, to put it differently, that âmaximum payoffâ or âNash equilibriumâ is no more or less a source of prominence than âhighest buildingâ or âonly bridgeâ; that being the case, we might as well study prominence or salience directly. And discerning prominence or salience often involves intuition, guessing, or even whimsy more than logic and calculation. If this is game theory, it might be called game theory for romantics. In Schellingâs words, âPoets may do better than logiciansâ and the game is âmore like âpuns and anagramsâ than like chess.â5
While Schelling does not purport to justify or derive focal points, he does describe some qualities that tend to make something focal: uniqueness; the state of being the status quo or the status quo ante; the suggestion of a mediator; fact-finding (which fills a âvacuum of indeterminacyâ); physical prominence; conventional priority (âheads or tailsâ gives a certain priority to heads); and analogy to a case recently brought to mind.6 The status quo criterion threatens to collapse the coordination game altogetherâwe have no coordination problem if everyone carries on as usualâthough Schelling brings up interesting examples of when there is a problem and âstatus quoâ by some description solves it. The other criteria are crude in the sense of being somewhat desperate, the kind of thing we grasp at for want of a more rational or efficient standard for agreement. As Schelling puts it, we resort to focal points when we seek an âexcuseâ that will let us come together rather than a âreasonâ for doing so in a particular way.7 Desperate people who (wrongly but not unreasonably) suppose things can get no worse than the lack of a monarch might pick an absolute monarch on the basis of height alone.8
Coordination problems have become the subject of a comprehensive literature and one which has further explored and specifies what the problem consists in. Robert Goodinâs definition stands out in scope and clarity: âA âcoordination problemâ is defined as existing whenever it is rational for all agents involved to prefer joint to independent decision-making.â Such problems obtain when two things hold. The first is âmutual involvementâ of actors (so that each actorâs welfare is affected by othersâ decisions, giving each an incentive to coordinate his or her decision with the rest). The second is that âindependent decision-making must involve risks of disagreeable outcomes for everyone involved.â This condition is not trivial: â[s]ometimes [actors] gain from the confusion accompanying independent decision-making and would prefer to keep the independent decisional processes.â9 (In the next chapter we shall see how this looks: literally, medieval.) When no common interest in joint decision making exists, no coordination problem exists: a group might then be coerced into common action, but no agent will seek it spontaneously. One should add that when such interests exist but are not perceived, a coordination problem may objectively exist but will not be addressed, much less solved.
When coordination problems do exist and are known to the actors, their hallmark is this obvious and mutual interest in common decision. In so-called cooperation games (prisonerâs dilemmas and their like; whether to pay oneâs taxes is the most accessible example), a selfish actor has an interest in doing something different from the others: i.e., my first choice is that others cooperate while I defect.10 In coordination games, that is not the case. I have no interest in fooling everyone into driving on the right while I drive on the left, nor in persuading those around me to speak a language different from mine. On the contrary, the motto of coordination is (as Goodin points out) âthe more, the merrierâ (corollary: âwhen in Rome, do as the Romans doâ)11âalso a theme of the next chapter.
Circumstances of Authority and Qualitative Coordination
Most literature on coordination is technical. It posits payoff matrices, deduces optimal strategies, and turns coordination problems into âgames.â For current purposes, such precision is not just unnecessary but misleading. The cases covered in this book, which involve solving political coordination problems through long-term and large-scale conventions, violate the technical assumptions of the game theorists not slightly or incidentally but deeply and essentially. Humean conventions are relevant, or most relevant, precisely where the simplifying assumptions of game theory, fully justifiable in many other circumstances, are deeply and essentially misleading.
Those who know a little bit about game theory typically think of it as involving agents who are choosing strategies as a means to achieving the largest possible payoff expressed in points (or units of happiness or utility), given that other players are trying to do the same. This picture involves two assumptions more stringent than strictly necessary: cardinal utility and instrumental rationality. Cardinal utility is fiercely controversial, for reasons too complex to cover here, except in a longish note.12 For now, I shall merely say that it is of little relevance for current purposes. Hume himself never used a cardinal terminology (which was arguably invented by Bentham, who also coined the word âmaximizeâ). When he spoke of utility it was in the sense of somethingâs being, broadly speaking, to everyoneâs advantage.13 And on substance, the circumstances of high constitutional politics do not lend themselves to cardinal precision. While the distributional consequences of authority conventions are crucially important, they are best captured as debates over political principles (as in chapter 7) rather than calculations over precisely known expected outcomes.14
As for instrumental rationality, the assumption that agentsâ choices are causally related to a future state that they prefer, both history and Humeâs History reveal that this assumption is sometimes violated, whenever agentsâ motives are intrinsic, expressive, or categorical. (As mentioned earlier, the ideal-typical assumption that agents are instrumentally rational is explicitly intended by Weberâand implicitly practiced by Humeâas a method for fully discerning and understanding the instances in which they are not.) Though Hume for the most part hoped for the triumph of instrumental rationality, he admitted that the mixture of religion with faction often brought about âunaccountableâ events in which âeffects correspond less to their known causes than is found in any other circumstance of governmentâ (H 5.67; compare 1.380, 3.206â7, 435â6; 4.18, 221; 5.342, 380, 493, 514; 6.3â4, 113, 120, 128â9). More rarely, fierce non-religious passions such as those for money, sex, revenge, honor, and power cause the most strategic actors to act non-instrumentally as well (see, e.g., H 2.153, 371, 385, 390, 2.433, 449, 463ff., 2.489; 3.128, 227, 242, 261, 321, 462, 478; 5.509; 6.275, 287 [note m]). In fact âa man wholly interested is as rare as one entirely endowed with the opposite qualityâ (H 2.489),15 though fortunately the strongest anti-instrumental passions tend to decay over time, whereas âthe sense of interest maintains a permanent influence and authorityâ (H 2.411).
The most ecumenical definitions of game theory, however, those meant to appeal to those skeptical of its relevance, do not posit cardinality and instrumentality and speak merely of preferences over states of affairs. In Eerik Lagerspetzâs formulation, game theory assumes only three components and two irreducible assumptions. The components are â(i) a specified set of n players; (ii) n sets of strategies, one for every player; (iii) n sets of rational preference orderings one for every player.â The assumptions are rationality (âplayers make their choice according to their preference orderingsâ) and full information (both the components [iâiii] and the rationality assumption are âcommon knowledge,â that is, all the players know who the players are, what they want, and that they are rational; and each knows that others know, and that the others know that they know that others know, and so on).16
But even these specifications and assumptionsâoften eminently reasonable when the actors are buyers and sellers bargaining in a marketâmay be questioned when it comes to the fundamental issues of political order and authority central to Humeâs History. Agents may have little idea who the relevant players are or what they might want, and in fact who the players turn out to be may be endogenous to the gameâs outcome. That is, which authority conventions each agent comes to live by will determine which other agents will be politically relevant in the future: the polity could expand or contract; foreigners might intervene or refugees leave, and the choices made in a game of authority could greatly affect, through the choice of political and social institutions, the number of players who are born, or survive, to play future rounds of the game.17 And agentsâ preferences may not be stable over time or predictable to themselves. In fact one way political conventions might work, as noted in chapter 2, is by reshaping some actorsâ preferences so that future adherence to conventions becomes easier for them (or their descendants).18 The advantages of conventions of authority take the form of peace, which puts out of business whole orders of society whose business is war; the partial privatization of religious belief, which undermines the claims to universal moral and political authority that all clerical factions once regarded as beyond question; and long-term economic growth, which is guaranteed to transform existing mores, outlooks, and aspirations beyond recognition, and to make possible achievements in art and science that cannot possibly be anticipated before they happen.19 Under such circumstances, a model that posits determinate actors with fixed preferences over outcomes may remain technically possible but is unlikely to illuminate the historical dynamic. The founders of states are chasing dim visionsââcommerce,â âfraternity,â the âgeneral welfare,â the âblessings of...