1 1 Abstract Rational choice as process: the uses of formal theory for historical sociology The major drawback of rational choice theory is not its individualistic approach, on the contrary it is well suited to explain aggregate outcomes; rather it is its essentially static, or at best cyclical, character which prevents it from coming to terms with social processes in which not only variables, but in the course of time also parameters and even 'constants' must be considered as changing entities. But this is incompatible with the requirements of formal conceptualization and statistical testing procedures. Nevertheless, in decisive episodes, human beings tend to be 'alert' and 'scheming': the key notions of rational choice theory are too productive to be ignored by historical sociologists who do well to incorporate them as intellectual concepts in a pragmatic manner. Abram de Swaan Amsterdam School of Social Research Rational Choice as Process The Uses of Formal Theory for Historical Sociology1 'La structure, c'est les autres'2 Sociology studies people in the relational networks they constitute. This constellation of interrelated human beings is both structured and changing: it is a structured process.3 Clearly, within sociology tensions exist between an approach that starts out from individuals and one that takes its point of departure in the constellation they constitute together; and, equally, there are strains between a viewpoint Thanks are due to Nico Wilterdink and Johan Heilbron, to Wouter Gomperts and other members of the 'psycho- and sociogenesis club', and to the many colleagues who participated in discussions of earlier versions in various seminars. A different version of this paper appeared earlier with contributions by Siegwart Lindenberg and Johan Goudsblom and a rejoinder by this author in Dutch in Amsterdams Sociologisch Tijdschrift 22.4, March 1996, and in English in The Netherlands Journal of Social Sciences 32.1, November 1996. 1 Structure is the others; after Jean-Paul Sartre on hell: 'l'Enfer, c'est le autres' (Huis Clos). 2 These formulas correspond closely to the concept of 'figuration' in the sociology of Norbert Elias; cf. What Is Sociology?. London : Hutchinson, 1978. 3 1 1 that highlights a state of affairs and, on the other hand. a perspective that focuses on change. Indeed, these are the two dimensions that constitute the cross on which sociology suffers its major dilemmas. The various schools in sociology may be situated along these axes. Rational choice theory unambiguously belongs in one quadrant of this cross: it opts for an individualist, even a reductionist strategy, and a static, at best a cyclical 1 FIGURE I dynamic │ │ ..................│..historical neo-evolutionist │ sociology theory │ : figurational │ : sociology psychoanalysis │ : │ : │ : │ : ─────────────────────┼───────────────────── individualist │ : structuralist │ : │ │ rational │ network choice │ theory theory │ │ │ │ static approach.4 Historical sociologists, on the other hand, may at times take a static view of the past and concentrate on individual actions, but by and large they gravitate toward a structural and a dynamic perspective: a process approach which is For a wholly uncompromising formula, a quotation from James Coleman, Individual Interests and Collective Action. Cambridge etc.: Cambridge U.P., 1986, p. 1: 'If an institution or a social process can be accounted for in terms of rational actions of individuals, then and only then can we say that it has been explained.' (and a little further along, p. 16:) 'I will start with an image of man as wholly free: unsocialized, entirely self-interested, not constrained by the norms of a system, but only rationally calculating to further his own self-interest.' 4 1 especially characteristic of Norbert Elias's figurational sociology.5 In a third quadrant one might find a school of theorizing that is at once static and structural: network analysis.6 The fourth quarter is reserved for the individualist and dynamic approach that is common to such divergent schools as psychoanalysis7 on the one hand and neo-Darwinian or geneticist evolution theory on the other.8 This strain has been taken up by social scientists who have tried to explain the evolution of cooperation from individualist and maximizing principles.9 For an introduction to historical sociology: Philip Abrams, Historical Sociology. Somerset: Open Books, 1982; Theda Skocpol (ed.), Vision and Method in Historical Sociology. Cambridge: Cambridge U.P., 1984; Denis Smith, The Rise of Historical Sociology. Cambridge: Polity Press, 1991. 5 For a programmatic collection: Barry Wellman and S. D. Berkowitz eds., Social structures: A Network Approach. New York: Cambridge U.P., 1988; for an elementary introduction: John Scott, Social Network Analysis; A Handbook. London etc.: Sage, 1991. 6 For an introduction: Sigmund Freud, Civilization and its Discontents in The Standard Edition of the Complete Works of Sigmund Freud, vol. 21, pp. 57 sqq. See also: Fred Weinstein and Gerald M. Platt, Psychoanalytic Sociology; An Essay on the Interpretation of Historical Data and the Phenomena of Collective Behavior. Baltimore/London: Johns Hopkins U.P., 1973. 7 For a popular but very radical introduction, Richard Dawkins, The Selfish Gene. London/New York: Granada, 1978 and by the same author: The Blind Watchmaker. New York/London: Norton, 1986. For applications of evolutionary selection to social institutions: //Dunn Hall, Allan Carling// 8 Cf. S.A. Boorman and P. R. Levitt, The Genetics of Altruism. New York: Academic Press, 1980; Robert Axelrod, The Evolution of Cooperation. New York: Basic Books, 1984; Robert Boyd and Peter J. Richerson, Culture and the Evolutionary Process. Chicago: University of Chicago Press, 1985; Henk de Vos and Evelien Zeggelink, 'The emergence of reciprocal altruism and group-living: an object-oriented simulation model of human social 9 1 The subject here is the encounter of rational choice theory and historical sociology. The core problem of rational choice theory is in one word 'order' - how are human efforts coordinated given that individuals pursue their own interests?10 The core question of historical sociology is to explain how contemporary society emerged from its antecedents, and, more precisely, how the scale of effective coordination could expand from the few dozen members making up a band of nomads to multinational states numbering a billion or more and to global exchange networks of similar size, all in the relatively short time span of some ten thousand years, less than five hundred generations. Notwithstanding the vast disparity between the knowledge interests of the two schools there is some common ground and there have been some intellectual encounters. The main issues in this confrontation are presented here, without attempting to review the entire literature or discuss specific technicalities. First, rational choice theorists have presented a reading of historical sociologists in order to reconstruct the argument of the latter in terms of the individualist premises of the rationality model. What are the main compatibilities between the two paradigms and which major incompatibilities remain? Second, within rational choice theory, attempts have been made (..vervolg) evolution' Social Science Information, 33.3, 1994, pp. 493-517. "Surely there is no problem in the social sciences that is more important than that of explaining why people cooperate." Jon Elster, Making Sense of Marx, 1985, p. 366. 10 1 to overcome some vexing paradoxes by introducing more dynamic concepts, especially in the analysis of the prisoners' dilemma. To what extent have these revisions succeeded in bridging the gap that separates historical sociology and the rational choice model? Third, rational choice concepts have been applied within the theoretical framework of historical sociological studies.11 What is the most promising strategy for incorporating ideas from the rational decision approach in historical sociological work? These three distinct, but related questions deserve a brief discussion to point out the problems and the prospects of combining rational choice theory and historical sociology. Several authors have confronted historical sociological studies with their version of a rational choice model: Jon Elster wrote a lengthy and elaborate critique of Marx' writings from a methodological individualist viewpoint, attempting to salvage those parts of marxist theory that could be reconstructed in formal terms. His discussion of Marx' theory of class consciousness is especially instructive for the present purposes, as it is entirely framed in terms of an n-person prisoners' E.g. Siegwart Lindenberg, 'Social Production Functions, Deficits, and Social Revolutions; Prerevolutionary France and Russia' Rationality and Society, 1.1, July 1989; see also Edgar Kiser, Kriss A. Drass and William Brustein, 'Ruler Autonomy and War in Early Modern Europe' International Studies Quarterly, 39, 1995, pp. 109-138, or William Brustein and Jürgen W. Falter, 'The Sociology of Nazism, An Interest-Based Account' Rationality and Society, 6.3, July 1994, pp. 369-399. 11 1 dilemma.12 In an interesting essay on the 'figurational sociology' of Norbert Elias, Hartmut Esser has pointed out that its historical sociological method agrees with the major canons of methodological individualism.13 And, in fact, it is not the oftcriticized reductionism of methodological individualism and its close relative, rational choice theory, which makes for an unbridgeable difference with the 'figurational' sociology of Elias and other historical sociologists. On the contrary, rational choice models and models from game theory are especially well-suited to demonstrate the interplay between individual strategies and aggregate outcomes. Many concepts from game theory, such as 'zero-sum game', 'maximin strategy', 'free-rider' or 'prisoners' dilemma', have acquired a second life in everyday discourse precisely because they can convey so convincingly the dramaturgy of actors whose fate depends on the other's Idem, pp. 342-71. For another, interesting discussion of marxist concepts that goes beyond the tenets of rational choice, cf. Alan H. Carling, Social Division. London: Verso, 1991. 12 Cf. Hartmut Esser, 'Figurationssoziologie und methodologischer Individualismus', Kölner Zeitschrift für Soziologie und Sozialpsychologie 36 1984, pp. 641-66. Much the same point is made by H. D. Flap and Y. Kuiper, 'Figuratiesociologie als onderzoeksprogramma' [Figurational sociology as a research program] Mens en Maatschappij 54.3 1979, pp. 323-369. For a polemic between Esser and Arthur Bogner, a student of Elias: 'Bemerkungen zu Hartmut Essers Aufsatz "Figurationssoziologie und Methodologischer Individualismus",' followed by a comment from Esser and a reply from Bogner, Kölner Zeitschrift für Soziologie und Sozialpsychologie. 37.4 1985, pp. 800-7. 13 1 decisions.14 There is indeed a close correspondence between the notion of 'unintended outcomes' in rational choice theory and the concept of 'blind process' in historical sociology, just as there is a strong analogy between respectively the concept of 'external effects' in the former and of 'interdependency' in the latter. At the core of the rational choice approach is the notion of rational maximization, i.e. the choice of the most efficient means to realize a given end. Such optimizing behavior is culture- and timebound, as historians and historical sociologists have amply demonstrated. Nevertheless, in a weakened form the rational choice axiom is valid for most of historical time. In competitive situations where physical security, vital resources, or reputation are at stake, people will remain alert and attentive to possible advantages to be secured, damages to be avoided; they will scheme and plot to secure their position and avoid setbacks. The common sense grounds for this assumption are somewhat strengthened by the fact that when physical, social and moral survival are at stake, those who do lose out risk being eliminated from the competition altogether.15 For an accessible introduction to game theory, cf. Anatole Rapoport, Two-Person Game Theory: The Essential Ideas. Ann Arbor: University of Michigan Press, 1966, and idem, N-Person Game Theory: Concepts and Applications. Ann Arbor: University of Michigan Press, 1970, For an application of game theoretic notions to a variety of social situations, see Thomas C. Schelling's classic The Strategy of Conflict. New York: Oxford U.P., 1963. 14 For a discussion of this argument, cf. Douglass C. North, Institutions, Institutional Change and Economic Performance. Cambridge/NY: Cambridge U.P., 1990, pp. 19-22. 15 1 In anthropology and in the history of early and medieval societies, a different view often prevails: pre-modern people were strangers to the calculating and fore-sighted stance that characterizes individuals in the Modern Era in the West. If this implies that people in subsistence economies were and are not out to maximize their money income over time in the manner of contemporary entrepreneurs in a market economy, the observation is correct and a truism. But if the point is that people in nonmonetary economies are propelled merely by passion and hemmed in by norms, all the while unmindful of their interests, it is a caricature. From the beginnings of sedentary agriculture, and maybe even earlier, people have shown themselves quite alert and scheming, at least in matters of marriage and warfare. Warriors simply could not afford to allow the passions of fear and rage to overcome them during battle. Peasant families have always carefully arranged marriages with an eye to preserving and extending their land holdings as the family heritage. Where markets emerged, this alert and scheming stance spread with them.16 In other words, the historical sociologist does well to assume a degree of alertness and a calculating stance, but what it is that is at stake in any given episode varies with time and place, Cf. Th. Haskell, 'Capitalism and the Origins of the Humanitarian Sensibility', Parts 1 and 2. American Historical Review 90.2, pp. 339-61; 90.3, pp. 547-66, for an account of the 'market-oriented mentality' which may well be compatible with and even promote altruistic propensities. 16 1 and sometimes several games are being played at once, be it about power, property or prestige. But nevertheless a wide gap remains, notwithstanding the bridge-building efforts of Esser and others: historical sociology, and certainly Elias' brand of it, essentially presupposes a process theory, whereas rational choice theory and methodological individualism are incorrigibly static, at best passing off iterative and cyclical models as dynamic constructs. At bottom, formal theory implies a ceteris paribus assumption: whatever is not explicitly introduced as a variable is assumed to remain constant. The process perspective, on the other hand, may best be introduced with a metaphor: the analogy of the clockwork with wheels that turn at different speeds; some cogs - turning so slowly that the casual observer hardly notices their movement nevertheless set in motion other wheels that move at visible speed and in turn drive a sequence of cogwheels at an ever faster speed, the last one connected to a balance wheel that by its frenetic oscillation controls the movement of the entire machine.17 Thus, within the very slow clock of biological time turn the wheels of human history and within it again, much faster, the gears of succeeding generations and finally the cogs of human actions and events that together control the pace of Cf. F. Braudel, 'History and the Social Sciences' in: On History. (transl. from the French) Chicago: University of Chicago Press, 1980, pp. 25-54, on 'durée'. For a discussion of key temporal concepts in historical sociology, cf. Ron Aminzade, 'Historical Sociology and Time,' Sociological Methods and Research 20.4, May 1992, pp. 456-480. 17 1 history. There is the rhythm of the species, the rhythm of societies, the rhythm of individual lives and of particular activities. But this is where the metaphor stops, since in last analysis, all clocks are cyclical and periodical, no matter how long the cycle may take, while biological evolution and human history within it occur only once.18 In formal theory construction, only a few quantities are treated as variables, most others are defined as parameters, held constant for the sake of the analysis, while still other entities do not even enter the argument as quantities that might change over time. For example, price indices are calculated from the prices of a 'basket' of goods and services; while the prices are expected to fluctuate, the composition of the basket is treated as a parameter, assumed to remain unchanged for the time being. Of course, price indices can be recalculated over a longer time span with a somewhat changed composition of the basket. But, over an even longer time period, say the past century, the proportion of goods produced at home versus that of goods bought in the market has changed gradually but profoundly. In other words, the share of the commodity basket in the family household has increased as the money economy expanded. What appeared as a constant for the short-term analysis turns out to vary in the long run. Moreover, in this time perspective a new problem presents itself: how to measure the proportion of goods and Cf. my The Management of Normality; Critical Essays in Health and Welfare. London/New York: Routledge, 1990, p. 5. 18 1 services in the household that were not purchased with money. Such developments usually remain entirely outside formal models, but they often do find their way into historical sociological theories of the process type.19 Rational choice theorists do not eschew such 'slow moving variables' out of a lack respect for history, but because additional variables make it so much more difficult to infer a limited set of precise predictions, and because uniform data that may be statistically processed become scarcer as time periods extend backward. As data become scanty and predictions fuzzy, statistical tests lose their bite and formal theories their discriminative edge. And yet, once formal arguments had served to identify new problems, time and again a breakthrough was achieved in rational choice and game theory, by next relaxing some formal restriction so as to generate a more realistic solution. T.C. Schelling chose to open up the argument to the total perception and foreknowledge of the players, their 'Gestalt', and this allowed him to suggest very convincing solutions to tacit bargaining games, especially games of coordination.20 Gary S. Becker, A Treatise on the Family. Cambridge, Mass. and London: Harvard U.P., 1981, pp.94-8, in discussing the 'quality of children' over long historical periods presents formal expressions that take into account both commodities purchased in the market and goods or services produced in the household by introducing time spent on child care as a variable, without, however, operationalizing the latter term. 19 20 Thomas C. Schelling, op. cit. 1 There is yet another consideration of major importance: People maximize along different dimensions that can not be reduced to a single denominator (or sequence). Rather, the incommensurability of various human strivings seems to be at the core of the human condition. In each 'existential choice' people decide what at that point prevails for them: honor or money, power or fame, learning or riches... No formula can a priori determine that choice. A theory that allows to reduce all strivings in human society to a single preference order is still as far fetched as the Grand Unified Theory is in physics, a single equation that should express all forces in the universe. Physics can do without such a construct, on weekdays. But due to this very shortcoming, rational choice theory is time and again pushed into irrelevance, tautology or indeterminacy. It is as great a defect of rational choice theory as the ingrained statical character that is the main theme of this article. Some of the most vexing problems of human cooperation may elegantly be expressed as dilemmas of collective action in terms of the rational choice model. Ever since Mancur Olson,21 social scientists have been painfully aware of the paradoxes in the analysis of collective action, i.e. actvities aimed at bringing about or maintaining a collective good. Such a collective good is Mancur Olson, The Logic of Collective Action. Cambridge, Mass: Harvard U.P., 1965. 21 1 'non-excludable': for technical or economic reasons, no one (within some social circle) can be excluded from using it. Moreover, in its purest form, that is as a 'public good', the condition of 'jointness of supply' applies: use of the good by one person does not diminish what is available for use by others. Finally, it is usually assumed that the efforts of a single person are not sufficient and those of all persons (within some social circle) are not necessary to bring about or maintain the collective good. A clean environment provides the classic example of a pure collective, or public good. Under these conditions, the suspicion that others may not join in the effort to create or maintain the collective good is sufficient for anyone to refrain from collaborating. As a result, it will not be brought about, leaving all participants worse off than they would have been if they had joined the collective action that could have produced it. The predicament is more acute, the larger the group concerned.22 The situation may be rendered as an n-person prisoners' dilemma game. From early on, it appeared that this paradox was in part the result of a completely static, one-shot analysis of collective action. The introduction of sequences over time had provided a solution in the case of - non-cooperative - zero-sum games without a saddle point: these games turn out to possess a stable This inference has been formally refuted and contradicted with a wealth of empirical argument by Gerald Marwell and Pamela Oliver, The Critical Mass in Collective Action; A Micro-social Theory. Cambridge: Cambridge U.P., 1993, pp. 38-59. 22 1 solution when considered as instances in an infinite sequence of iterations in a 'super game': the 'mixed strategy' of random choices in a fixed proportion.23 Michael Taylor applied this approach to another class of games, prisoners' dilemma games with their especially vexing paradoxes, and succeeded in demonstrating that the 'supergame' consisting of an infinite sequence of prisoners' dilemmas, under certain conditions does possess an equilibrium outcome based on cooperative strategies.24 Although the analysis was hardly dynamic, at least a notion of time was introduced in the analysis, a concept of repetition, of actors' anticipation and of their discounting future outcomes. In his final comments to the argument, Taylor remarks: Perhaps the most important shortcomings of the Prisoners' Dilemma supergame as a model of the process of public goods provision is that it takes place in a static environment: the supergame consists of iterations of the same ordinary game. In some of the public goods problems of interest here, a more realistic description of reality would require a changing payoff matrix, possibly a changing set of available strategies, and even a changing set of players.25 Cf. John von Neumann and Oskar Morgenstern, Theory of Games and Economic Behavior. Princeton, N.J.: Princeton U.P., 1944. 23 A very elegant argument based on sequences of bilateral encounters in a population is presented by Axelrod op. cit.: Cooperation unrequited again leads to loss, but when reciprocated it results in mutual gain. A 'tit for tat' strategy of initiating every interaction with cooperation but punishing the opponent's refusal to cooperate with immediate non-cooperation turns out to be most advantageous individually and might in due course be adopted throughout the population. The encounters are reiterated on the same terms, but in the population as a whole; in this case a process does occur: the spread of the 'tit for tat' strategy of initial cooperation and prompt revenge for non-cooperation. 24 Michael Taylor, Anarchy and Cooperation. London: Wiley, 1976, p. 96. Taylor continues with an interesting suggestion (p. 25 1 Elster, commenting on Taylor's analysis, suggests that after prolonged interaction the actors will begin to form a stable group and each individual's utility function will come to reflect the welfare of the other players. In other words, the individuals will begin to care for each other and what used to be the gains of individual defection will not look so attractive anymore, once the damage done to others is also taken into account. The final result is an 'Assurance game' which has a - conditionally cooperative solution.26 Unmistakably, Taylor's afterthoughts and Elster's reflections refer to an approach that goes beyond mere iteration, allowing for parameters to adopt new values, i.e. they incorporate changes over time and in a certain direction, in other words, the authors begin to consider collective action as a process. In Michael Hechter's Principles of Group Solidarity,27 collaboration is the result of members' dependence on the collective good and the group's capacity to enforce compliance: (..vervolg) 97): 'Where, for example a "commons" (--) is being exploited, the payoffs might decrease steadily as more and more non-Cooperative ("exploitative") choices are made over time...' Jon Elster, Ulysses and the Sirens. p. 146. Cf. also his discussion in Making sense of Marx, Cambridge/Paris: Cambridge U.P./Maison des Sciences de l'Homme, 1985, pp. 358-367. 26 Michael Hechter, Principles of Group Solidarity. Berkeley etc.: University of California Press, 1987; see also Michael Hechter and Satoshi Kanazawa, 'Group Solidarity and Social Order in Japan' Journal of Theoretical Politics 5(4), 1993, pp. 455493, 27 1 its 'control capacity,' which rests on the possibilities to monitor the members' behavior and the availability of selective sanctions. In his subsequent argument, Hechter relaxes some of the most stringent assumptions of rational choice theory: first of all, he is willing to face the fact that people have been socialized to cooperative ('pro-social') behavior in their past (they have a past!); secondly, he considers the possibility that the group itself has a history in the course of which it acquired its capacity for control. Hechter provides no conclusive answer to the question how such control mechanisms may emerge in groups, but it clearly is the result of deliberation and interaction among group members who are equipped with memories, e.g. of prior socialization, and with knowledge of the external world, e.g. of the norms that exist out there. Moreover, a concept of process is central but mostly implicit in the argument. Some people may profit much more from the supply of a collective good than others, some people have more time or money available to spend on collective action: those with the strongest interest and the largest resources will be the first to contribute towards the collective good, and if together they are effective in bringing it about they constitute the 'critical mass' for collective action.28 In this approach, developed by They need not even explicitly collaborate towards the collective good: people with similar 'social production functions', each on their own, may act to prevent major loss, e.g. rebel against tax legislation, in 'tacitly coordinated, individual collective action.' Lindenberg, op.cit., pp. 51-77. 28 1 Marwell and Oliver, the sequence of decisions is of the essence, and so is the 'production function' that defines how the supply of the collective good increases with each additional resource contribution. Clearly, decisions are interdependent and collective action occurs in a sequence over time, albeit within strict parameters.29 Using ideas from network theory, first advanced by Granovetter30, Michael Macy has proposed a simulation model in which participants take their cues for cooperation from others in the network according to the strength of their ties. After several rounds of costly inactivity the participants may learn that cooperation can be more rewarding and that others think so too. Again, once a critical mass has been achieved, a stampede towards collaboration may ensue.31 (..vervolg) Marwell and Oliver, op. cit., p. 15: 'We assume that each individual's interest and resources are fixed characteristics at the moment of decision about collective action. Violation of this assumption would necessitate rewriting our models. Our models cannot capture the dynamics that arise when people's understandings are changing over time. We do not believe that interests and resources are actually static, but we do believe that the patterns we can identify using static models are valuable baselines that must be constructed before we can hope to develop models that successfully capture complex dynamic processes. 29 Mark Granovetter, 'Threshold Models of Collective Behavior', American Journal of Sociology 83, 1978, pp.1420-43. 30 Michael W. Macy, 'Chains of Cooperation: Threshold Effects in Collective action' American Sociological Review 56, 1991, pp. 730-47. 31 1 What appears as an afterthought or as a latent theme in the preceding analyses is the explicit point of departure for the theory of the collectivizing process: collective action is considered as a process which produces both the collective good and the collectivity with respect to which the good is collective.32 In the course of their effort to establish some collective good, as the interaction proceeds, the actors gradually come to constitute a group, with a 'we-feeling', with shared memories, expectations and values. In other words, the capacity to monitor and sanction, to control members' compliance, is a property of the collectivity that may emerge in the process of collective action. Moreover, in this perspective, the dilemmas of collective action are themselves characteristic of a transitory phase. They occur when participants are interdependent, and aware of their interdependence, but without their actions yet being coordinated at a higher level of integration: the level of the collectivity. Once collective action does get underway, it may bring about both a collectivity and a collective good at this next higher level of integration. 'The main fault of formal theory as it stands is its inability to deal with processes - with changes over time. This reduction from dynamics to statics may be overcome in a sociogenetic approach. The logical paradox disappears in sociological analysis: a paradigm shift.' Cf. the introduction to my In Care of the State; Health, Education and Welfare in Europe and the USA in the Modern Era. Cambridge/New York: Polity Press/Oxford U.P., 1988 , which sets out to apply the concept of a 'collectivizing process' in a comparative historical sociological study of the emergence of welfare states in the long run. 32 1 But in order for the collective action to begin, somehow the initial, reciprocal suspicion of parasitism (or 'free-riding') must be overcome. This may be achieved by 'entrepreneurs' who manipulate members' expectations into believing that the others will cooperate; or it may come about because members hold illusions, underestimating the costs of cooperating while others do not, or overestimating the others' readiness to join the effort; or it may be brought about by external interventions such as grants and fines or coercive measures. Once collective action is initiated, it may bring about a collectivity which is capable of imposing selective incentives. The cheapest of these are informal sanctions such as ridicule, rebuke and gossip, which may well come at no cost or even at a premium: many people at least occasionally enjoy the opportunity to scold or blame others for their delinquency, especially if one's peers agree.33 But for such sanctions to be effective, actors must be in a position to communicate freely among themselves and to assess one another's contributions (what Hechter calls 'monitoring'), and there must be some agreement In the literature it is routinely assumed that such disciplinary measures are imposed only at a cost and that this effort again evokes dilemmas of collective action, and so forth, into infinite regression. Cf. Gerald Marwell and Pamela Oliver, op. cit., p.8: 'The problem is simply that somebody has to pay for the selective incentive, and paying for a selective incentive is also collective action in that it will provide a benefit to everyone interested in the collective good, not just the people who pay for the incentive.' But the selective incentive for some may be a pleasure to administer. 33 1 among participants as to the distribution rule which has been violated. As the collectivizing episode proceeds and collective action continues, an emergent collectivity may acquire all sorts of resources for controlling its members. Unilateral defection may no longer appear so tempting, and indeed, actors might even begin to care for one another, allowing someone else's welfare to enter their calculations. This conception of collective action is a long way from the formal and static models presented in pure rational choice theory. However, it also fall short of a theory that allows precise predictions to be inferred from its assumptions. Moreover, it is unclear so far which body of empirical data the theory refers to.34 Of course, formal rational choice theory does not always yield precise or unique predictions either, and when it does they often turn out to be wrong, whereas the empirical referents of the theory are often equally undetermined.35 But in Cf. Hechter, op. cit. p. 55. 'Often the data required to test this kind of theory do not exist. Perhaps if enough researchers become intrigued with the theory, systematic attempts to collect the relevant evidence may be made. But the lack of suitable ready-made data hinders the appeal of any new theory.' 34 The rational choice 'literatures' in American political science have hardly produced testable hypotheses, and of those, very few have been actually tested against empirical data, almost none was supported by the facts; cf. Donald P. Green and Ian Shapiro, Pathologies of Rational Choice Theory: A Critique of Applications in Political Science. New Haven: Yale U.P., 1994. 35 1 more general terms, formal and static rational choice theory yields a gain in precision and systematics, at the price of a loss in scope and relevance, while process theories offer the reverse trade off.36 The most promising strategy for assessing the validity of process theories of collective action is to reconstruct carefully specific collectivizing episodes from the available evidence and determine whether such empirically grounded accounts lend plausibility to the assumptions. This is a far cry, indeed, from the ideal of a definitive test of the theory's inferred predictions against a body of uniform data with the aid of standard statistical procedures. But process theories, operating with shifting terms, pose other requirements of validation. They are best tested in the course of a discursive reconstruction of events, i.e. in a historical sociological account. The process theory itself should contain the concepts necessary to organize the narrative and define the domain of relevant events that serve to test the theory's plausibility.37 The four criteria are proposed by Johan Goudsblom, Sociology in the Balance; A Critical Essay. Oxford: Blackwell, 1977, pp. 2-5, 196-8. 36 Edgar Kiser and Michael Hechter, 'The Role of General Theory in Comparative-Historical Sociology', American Journal of Sociology. 97.1, July 1991, pp. 1-30, (p. 11), defend general theory on the one hand against the 'historicist particularists' who 'conceive of the relation between particular events as contingent and accidental' and on the other hand against the 'inductive generalists' of comparative-historical sociology who lack clear criteria for defining the class of relevant events. 37 1 Taylor and Axelrod were both able to demonstrate that cooperative strategies might be stable under certain conditions by treating each game as a round in a long sequence of iterations. Taylor, Elster and Hechter all suggested that the dilemmas of collective action might be resolved if some crucial parameters are treated as variables, if payoffs are allowed to change in the course of the sequence, if players may be added or removed and if available strategies are introduced or eliminated in the course of interaction. This transforms the supergame of consecutive but identical games into a sequence of gradually changing games, i.e. into a process. In this process perspective, the capacity to control the actors, to effectively coordinate their efforts, is an emergent property that results from ongoing collective action. In the course of a collectivizing episode a collective good is realized, and at the same time a collectivity that coordinates the individual actions into an increasingly coherent effort. In this manner mutually interdependent actors, who are aware of their interdependence, may overcome their dilemmas and come to constitute a collectivity at a higher level of integration.38 Of course, the 'actors' may represent individuals, or families, or villages, or towns, corporations, states etc. At some point, the question may arise whether and why the entity that functions as an actor is indeed a unit that effectively coordinates the actions of its constituent elements. Rational choice theorists try to make good on their promise to answer that question all the way down to the level of individual persons. But the question why aggregate agents are coherent need not always be answered first. E.g., the interactions of states in world society may be studied by treating them as coherent actors, pending a full account of their internal composition, coherence and consistency. 38 1 It may be possible to render such collectivizing episodes in the formal terms of a dynamic rational choice model, but available mathematics do not adapt themselves easily to such process constructs, due to the poverty of formalism: formal models of process theories are rare or nonexistent. Although historical sociological theories will not soon be couched in the formal terms of the rational choice model, process theorists may borrow its central concepts and adopt notions from game theory or the theory of collective action to guide their research and structure their narrative. If indeed, the central question of historical sociology is how the scale of effective coordination could expand from an order of magnitude of dozens to billions in the span of a few hundred generations, then the problem of collective action should be at the heart of historical sociological theory: the dilemmas of collective action occur, as mentioned before, precisely in the transition from one level of integration to the next. In order to explain this transition, the question must be answered why people do collaborate and why they do not defect. Once the collectivity is in place, and payoff matrices have been modified accordingly, cooperation may be taken for granted and other questions take precedence, e.g. the process of integration at the next higher level. In other words, the analyst may take a structural view in dealing with one level or one stage of integration and a conflict theoretical, rational choice perspective at the next higher, more recent, still emergent level. 1 In this perspective, historical sociology should make pragmatic use of rational choice theory, adopting its key notions as search strategies and 'sensitizing' concepts. As formal theory is constitutionally incapable of dealing with processes, there is no reason why historical sociologists should swallow its precepts whole and allow it to dictate their standards to them. Nor should they let their research questions be determined by the consideration whether sufficient and uniform data are available for statistical testing. Rather, in this instance also, the questions should be generated by priorities of theoretical relevance, while research and validation methods are to be adapted to pragmatic considerations. Rational choice theorists identify their professional ideals with those of the exact sciences, with population biology, mathematical micro-economy, and experimental psychology, and that is from where they draw their standards of inference and validation. Historical sociologists, on the other hand, do not seem to attach equal importance to logical inference and statistical testing, but neither do they simply adopt the criteria of history or the other humanities: beyond the precision of historiography they aim for the scope and systemics that grand theory claims for itself. This double ambition often locks them into a double bind of having to satisfy the frequently contradictory criteria of both history and science. It is this predicament that defines the historical sociologists' habitus in the field between 'history' and 'social "science".' 1 But what most deeply separates rational choice theorists and historical sociologists is the tendency of the former to think in terms of steady states versus the predilection of the latter for a process perspective. Historical sociologists must deal with the increasing scope of human coordination as their core problem and they do well to assume that people display a degree of scheming alertness during critical episodes, a stance that may well have spread and intensified in the course of history. For these reasons, historical sociologists will profit greatly from using rational choice concepts and using them as they see fit. If one had to explain the course of a conflict, one should always look first at the kinds of behavior that in a given figuration maximize peoples' chances of power, success and prestige. One certainly ought to look carefully at the hidden gains of apparently impulsive and emotional behavior; one should trace how people ostentatiously stick to their principles and in the meantime estimate the odds that others will do the same; and if people stick to their habits, one should want to know if that is their way of hedging against loss from changing circumstance. In brief, it is necessary to look for the alert and scheming components of behavior in interaction. Why? Because they are so effective. All along one must avoid the premature quantifications, circular arguments and tautologies that rational choice theory is apt to produce as soon as it transcends the economic sphere in 1 which ideally everything can be brought under the one denominator of money as the unit of valuation and calculation. As a matter of fact, the use of a rational choice model is by itself a choice model of a pure rational choice strategy: in an academic world dominated by the exact sciences, where economics in its turn dominates the social sciences, the rational choice model is most highly regarded and rewarded most generously. In conclusion, the core ideas of rational choice theory must be liberated from the scientistic straight jacket of 'modelled' theories, massive data sets and statistical tests. These core ideas may then be applied quite profitably to historical sociological theory in an evolutionist vein.