H Donald Forbes. Handbook of Political Theory. Editor: Gerald F Gaus & Chandran Kukathas. Sage Publication. 2004.
Positive political theory, narrowly understood, means rational choice theory applied to the study of politics. More broadly understood, as it will be here, it can refer to a much wider array of analytic approaches and final goals. Its limits are set by two familiar contrasts: positive, or what is, is contrasted with normative, or what ought to be; and theory, in the sense of abstraction and explanation, is contrasted with detailed descriptions of particular cases.
The aim of this chapter is to clarify what it can mean, in the academic study of politics, to give simplifying empirical analysis of some kind priority over plain description and explicit prescription. Three main kinds of positive theorizing will be distinguished, which will be called, for want of any better terms, conditional, rational, and intentional. Most of professional political science fits within one or more of these categories, but only a few examples of each can be selected for discussion here.
The analysis of political facts is often cast in terms of the relations between independent and dependent variables, and when it is, the aim is almost always to isolate some relatively simple functional relationships among the values of the variables. Do any of the independent variables correlate strongly with the dependent dimension? Are any of these correlations more than just correlations—that is, evidence of causal connections? What are the necessary and/or sufficient conditions for the outcomes of interest? More generally, which prior conditions make these outcomes more or less likely? In the present context, conditional may be better than the standard term causal for distinguishing the kind of causal analysis suggested by these questions (cf. King, Keohane and Verba, 1994).
Consider a simple example. Why do some citizens of Detroit vote Democrat and others Republican? Surveys may suggest that Catholics vote significantly more often for the Democrats than do Protestants or Jews. This correlation between religion and vote may be a clear statistical fact, derived by rigorous reasoning from some more elementary (or ‘brute’) facts about the way a sample of Detroit’s residents have answered questions about their race, religion, occupation, education, and so on. Admittedly, it may have almost no relation to what these voters would say if asked to explain why they voted as they did (they might say almost nothing about their religious backgrounds or beliefs), but it may still be an important fact in the context of practically oriented speculation about why voters really vote the way they do and what can be done to make them vote as one would like. In short, it may be part of a ‘causal’ theory of voting, in Detroit or elsewhere, and the theory may be true or false, regardless of what one thinks ‘normatively’ about voting for any particular party.
Rigorous statistical reasoning was first widely used in political science to establish and to explain or interpret simple relationships of this kind. The quantitative study of public opinion and voting is now one of the largest subfields of the discipline. Few of its many findings are perhaps of much interest to political theorists, but the methods and overall approach of such ‘behavioural’ research are another matter. And recently some of their most important applications have had to do with large claims about the causes and effects of democracy, as the following examples will show.
The Democratic Peace Hypothesis
Liberal democracies have rarely or never gone to war with each other. But can we say that democracy is a cause of peace or a sufficient condition for it? The democratic peace hypothesis is essentially the claim that wars have occurred and can occur only between autocracies or between democracies and autocracies. If all countries were democracies, there would be no wars.
The hypothesis can claim a root deep in modern political theory (Doyle, 1983; Cavallar, 2001; Franceschet, 2001). Whether true or false, it may have influenced policy-making at the highest levels. Both theoretically and practically, therefore, it seems important to determine whether it is in fact true or false.
The literature bearing on the hypothesis has grown dramatically in the past 20 years and has gradually become very technical. The earliest statistical studies (Babst, 1972; Small and Singer, 1976) suffered from some obvious shortcomings, but more recent studies have been models of careful conceptualization, assiduous data collection, and sophisticated multivariate data analysis. The basic challenge has been to justify a causal interpretation of a striking statistical regularity. To do so statistically one must introduce additional variables and test more complex models. Unfortunately, the more elaborate the statistical models, the more precarious their empirical foundations. War is a rare event, and since most of its causal conditions change only slowly, one cannot easily determine, from the examination of the annual data used in most statistical studies, whether there is any statistically significant relationship between the occurrence of war and particular background conditions, such as the presence or absence of democracy. The choice of an appropriate probability model is evidently a crucial first step in the analysis of the historical record, but it is very difficult to see which model provides the proper benchmark. Moreover, since the relevant cases are so few, the coding of one or two problematic ones (Spain’s status as a democracy in 1898, Finland’s status as an enemy of the Allied powers from 1941 to 1944) can have a substantial impact on the results of any statistical analysis.
Despite these difficulties, there is now a consensus that empirical research generally supports the hypothesis: joint democracy seems to be a sufficient condition for peaceful relations between states (for reviews of the literature see Chan, 1997; Ray, 1995; 1998; Russett, 1993; Russett and Oneal, 2001). This now widely accepted ‘empirical law’ about ‘democratic dyads’ provides an outstanding example of statistically based causal theorizing in political science.
Even strong and well-established statistical relationships invite conflicting causal interpretations, however. Thus Joanne Gowa (1999), using the same historical data as many other studies of democracy and war, suggests that there was a different relationship between these variables before World War I than there has been since World War II. Before World War I, it seems, democracies may have been more likely than autocracies to threaten each other militarily and no less likely to be involved in war. Only since World War II do the data support the idea of a ‘democratic peace.’ In other words, the hypothesis does not hold universally, according to Gowa, but only as a statistical rule in particular circumstances, as a by product of a particular structure of alliances. Other recent studies have advanced a related critique, suggesting that broad ‘cultural variables’ (similarities of interest and outlook) are more important than ‘structural variables’ (forms of government) in explaining the relations between states (Gartzke, 1998; Henderson, 1998; Kacowicz, 1995) or that other political similarities, such as joint republicanism or joint dictatorship, may be as strongly associated with peace between states as joint democracy is (Peceny, Beer and Sanchez-Terry, 2002; Weart, 1998; Werner, 2000).
Democratization and Ethnic Conflict
Controversial hypotheses can sometimes be defended by combining them with others in a more complex theory. Each hypothesis, in isolation, may be vulnerable to damaging objections, but combined with others, it may become part of a much sturdier web of belief. This possibility is nicely illustrated by recent discussions of a practically very important objection to the democratic peace hypothesis.
Mature democracies may not fight wars with each other, and they may have reliable ways of resolving their internal conflicts, but what about countries in transition to a more democratic form of government? The collapse of autocratic authority may mean the end of power-sharing arrangements between ethnic or national rivals. Moreover, the threat of majority rule may give traditional autocratic elites a motive for fostering ethnic or national strife, to block further democratization. Thus democratization, in the context of ethnic diversity and latent ethnic conflicts, may produce not peace, but civil and international war. Examples that seem to fit this pattern come easily to mind, but do they illustrate a general rule?
Attempts to deal with this question in a straightforward way, along the lines suggested by the studies cited above, have yielded inconclusive results (Ellingsen, 2000; Enterline, 1996; Gleditsch, 2002; Gleditsch and Ward, 2000; Hegre et al., 2001; Mansfield and Snyder, 1995; 1996; 1997; 2002; Mousseau, 2001; Thompson and Tucker, 1997; Ward and Gleditsch, 1998; Weitsman and Shambaugh, 2002). The relationship, assuming it exists, seems to be too weak to stand out clearly from the multitude of other relationships involved in the causes of war and domestic turmoil. In the rigorously quantitative literature, therefore, methodological disputes (about the shortcomings of different data sets, the definition and coding of variables, the treatment of ex-colonial regimes, the calculation of significance levels, and so on) have tended to divert attention from the basic idea that promoting democracy (in China, for example) may actually increase the risks of war. A practically important claim is lost from sight in a blizzard of methodological minutiae.
The merits of the new hypothesis about democratization and war proneness are clearer when it is evaluated in a more ‘qualitative’ way. Thus Jack Snyder uses a variety of historical case studies to suggest that ‘none of the mechanisms that produce the democratic peace among mature democracies operate in the same fashion in newly democratizing states. Indeed, most of them work in reverse’ (2000: 55). The sense of security mature democracies feel in dealing with each other, their commercial rather than military preoccupations, the aversion of their peoples to war and their unwillingness to bear its costs: these and other checks on warlike behaviour may all be overridden in semidemocratic regimes, where power elites, threatened by democracy, may foment war as a way of bolstering their power, where wealthy industrialists may profit from the preparations for war, and where ordinary citizens may be neutralized or led astray by unfair constraints on electoral competition, disorganized political parties, and partial media monopolies. Case studies to illustrate these possibilities fall on the ‘qualitative’ side of the standard quantitative-qualitative divide, but the goal of the qualitative research can remain, as this example shows, ‘quantitative,’ that is, the discovery of simple correlations between background conditions and a dependent variable of interest.
Snyder’s basic contention is that traditional autocratic elites create exclusionist ethnic nationalisms when their power is threatened by the spread of democracy. In a quite ‘rational’ way, they provoke nationalist conflicts to protect their own interests. By claiming to govern in the name of a threatened people, they avoid having to surrender real political authority to the average citizen. ‘Nationalist conflicts arise as a byproduct of elites’ efforts to persuade the people to accept divisive nationalist ideas’ (2000: 32). Nationalism is thus the intervening variable between Snyder’s regime variable (democratic, democratizing, etc.) and his dependent variable of internal or external violence. Different varieties of nationalism (three ethnic, one civic) mediate the postulated relationships between mature democracy and peace, on the one hand, and between democratization and war proneness, on the other hand. Snyder’s theory lifts these relationships out of their statistical context, one can say, and gives them a more understandable meaning, with a ‘theoretical logic’ illustrated by the case studies.
The resulting theory of nationalism is certainly plausible, and it has a distinguished pedigree, going back to early scientific studies of the deviousness of princes, but a really convincing demonstration of its merits would require a more systematic discussion than Snyder provides of the alternatives to it—the other causal theories that have been abstracted from the vast historical and social scientific literature on nationality and ethnicity. Nonetheless, the rhetorical strategy of the book—supporting a shaky statistical generalization and putative causal law by means of case studies and a ‘theoretical logic’ loosely related to the idea of rational individual choice—is effective. It resembles the one employed in another recent and very influential theory about the conditions of democracy or good government.
Social Capital and Democracy
Social capital has different meanings in different contexts. Here it will be used for the variable Robert Putnam (1993) argues is a powerful determinant of effective democratic government. This determinant is the number of ‘horizontal’ linkages between individuals of equivalent status and power in voluntary associations such as choral societies, soccer clubs, hiking clubs, birdwatching clubs, literary circles, and the like. The more such linkages in a region, the better was the performance of that region’s government, Putnam found in his celebrated comparative study of the 20 regions of Italy. The relevant correlations were amazingly strong, and they pointed to the conclusion that a dense network of voluntary linkages is a crucial condition for strong, stable, responsive, effective democratic government.
Putnam maintains that ‘social trust’ (which he also calls social capital) is the variable connecting associational density to democratic performance. Trust is vitally important for a society, he says, because it helps to overcome ‘dilemmas of collective action’ and thus ‘to solve the fundamental Hobbesian dilemma of public order’ (1993: 112). Trusting and trustworthy citizens are more able to co-operate with each other, on the basis of voluntary agreements, than are those who lack trust in each other and cannot make credible commitments. A dense (and closed) network of civic engagements sustains generalized trust because it threatens naturally self-interested individuals with realistic punishments for defecting from their commitments (Coleman, 1988; 1990). In looser, more open social networks, individualism or narrow self-interest (opportunism, free riding, etc.) is more likely to flourish, so that all must forgo many opportunities for mutual gain. Trust, and the norm of reciprocity associated with it, serve to reconcile self-interest and solidarity. They ‘lubricate’ co-operation, not just in politics, but also in economics. In short, ‘good government in Italy is a by product of singing groups and soccer clubs’ (Putnam, 1993: 176).
Since its publication, Putnam’s remarkably suggestive analysis has been exposed to a great deal of critical scrutiny. Some have objected to his depiction of Italian society and politics; others have challenged the application of his theory to other countries, particularly the United States. Putnam, for example, may not have paid sufficient attention to the role that the Communist Party of Italy played in creating good government (operationalized as pollution controls, daycare centres, responsive bureaucrats, etc.) in those regions where it was strong (Tarrow, 1996). Could the crucial independent variable have been, not singing groups and soccer clubs, but communist cells? And how many regions are there really, from a statistician’s standpoint, in Italy? Are there the 20 that are distinguished in law and that are the basis for Putnam’s statistics, or are there really just two distinct regions, North and South? The weight of the statistical evidence must evidently depend on the answer to this question.
Similar problems appear when the theory is applied to other countries. Some support for its general applicability has been found in studies of the American states, even though the relevant correlations are distinctly weaker (Putnam, 2000; Rice and Sumberg, 1997; Rice and Arnett, 2001). Other comparative studies are not so encouraging, however. Peter Hall’s (1999) detailed study of Britain suggests that changes in norms and trust over time may be unrelated to changes in the vibrancy of associational life. Susan Pharr (2000) and Donatella della Porta (2000) make strong cases for attributing high levels of distrust and dissatisfaction with politics in Japan and Italy respectively, not to changes in social capital (in the sense of networks), or to the performance of the economy, but simply to the conduct in office of each nation’s politicians (cf Jackman and Miller, 1998). A number of critics (e.g. Berman, 1997; Fukuyama, 2001; Levi, 1996; Varshney, 2001) have argued that different kinds of social capital may have different effects, so that democratic political performance may be threatened by its ‘bad’ or ‘uncivic’ forms, difficult to distinguish in principle from its more desirable forms. As noted above, a simple horizontal vertical (or secular-sacred) distinction seems to have worked well for Putnam in Italy, but it may not be so easy to apply and justify elsewhere. (In fact it is silently dropped in Putnam, 2000.) Even if civic norms and trust are consistently related to performance, associational activity may not be (Knack and Keefer, 1997). And heterogeneous communities, where ‘bridging’ social capital is most needed, may be the least able to develop it (Alesina and La Ferrara, 2000; 2002).
Putnam’s chain of correlations (from social networks through trust to democratic performance) gains its aura of causal necessity, not so much from the strength and persistence of the statistical relations he and others have been able to demonstrate, as from the reasoning about collective action problems that accompanies the presentation of the still somewhat scanty evidence. If the chain must hold in theory, one assumes, then surely its links must be observable in the facts.
Some Provisional Generalizations
The studies cited so far illustrate the maturing of the kind of positive science of politics that the partisans of the ‘behavioural’ movement in political science were calling for 50 years ago. The early behaviouralists could provide only vague outlines and very simple examples of the more scientific research that they thought should replace intellectual history and institutional description as the core political science disciplines (e.g. Easton, 1965; Easton and Dennis, 1968). Their opponents could reasonably argue that nothing coherent or worthwhile would ever come of their attempts to build ‘empirical theory.’ Impatient critics could wave away the whole enterprise, saying that it might serve to show how Catholics voted in Detroit, but not much else (Taylor, 1968: 90). Such high-handed dismissals are less effective today, where research workers in the social sciences have access to vast archives of machine readable data from scores of countries, and they routinely employ far more powerful methods of statistical analysis than were generally available even a generation ago. The embarrassingly nebulous grand theories of the recent past—systems theory, structural functional theory, group theory, and the like—have receded from view. Attention now focuses on demonstrable relationships between measurable variables of obvious importance, such as democracy and war, and their analysis does not stop with the establishment of a few simple correlations.
To be sure, studies of the kind cited above remain vulnerable to some common objections. The difficulty of operationalizing key concepts such as democracy, war, nationalism, and good government is obviously a source of serious problems. Such ‘essentially contested’ political ‘variables’ do not lend themselves to easy quantification, or even identification, for statistical analysis. In addition, a serious, often insurmountable source of difficulties is the complexity of the background conditions that may have to be untangled before any simple causal connections can be shown. A realistic statistical model of the phenomena of interest may involve many variables whose effects rebound on their causes, making statistical estimation extremely difficult. Nonetheless, statistically based causal analysis does not require for its justification that every statistical study make a major contribution to scientific knowledge or that it be beyond reproach. It requires only that there be rigorous ways of testing hypothesized relationships and untangling the webs of conditioning variables in which they are embedded. The data and methods used in a particular study may be inappropriate, but this will be shown by comparing its assumptions and results with those of other such studies, not by abandoning statistical reasoning altogether for some radically different way of establishing causal conditions. Even case studies, as Gary King, Robert Keohane and Sidney Verba (1994) have argued, can provide grist for the statistical mill.
The trend towards statistical reasoning is even more striking when it is viewed from a longer historical perspective. More than a century and a half ago, John Stuart Mill clearly explained the basic ‘logic’ of a positive social science in Book VI of his System of Logic (1843). He showed that there could be no fundamental differences between social and psychological inquiries (‘the moral sciences’), on the one hand, and the natural sciences, on the other, in so far as they were all ‘inquiries into the course of nature,’ that is, attempts to discover the background conditions that produce particular phenomena. Much of contemporary social science is directly descended from his philosophy. But Mill seems to have had no inkling when he wrote that the growth of statistics, not just as data and methods, but as a way of reasoning about cause and effect, would transform the character of the science he projected and narrow its concerns, making it almost indistinguishable from the quantitative analysis of social and economic policy.5
Mill saw more clearly that the ‘logic’ involved in developing and testing scientific hypotheses about sequences of conditions or events is quite different from that required by ‘Practice, or Art, including Morality and Policy’ (the title of ch. 12 of Book VI). Mill’s distinction between ‘science’ and ‘art’ is our familiar fact-value distinction: the statements of science are in the indicative mood, he said, while those of art are in the imperative or optative moods. (They have to do with defining and harmonizing our different ends or objects of desire, such as ‘health’ and ‘the happiness of mankind.’) This distinction has always been controversial, but in the present context it is easy to see how the ‘scientific’ investigation of causal conditions can be separated from the ‘ethical’ discussion of ends, and also easy, with Mill, to regard the two kinds of inquiry as complementary. Empirically grounded simplifications help political practitioners to know the conditions of the effects they seek and therefore, to some extent, whether they are worth pursuing.
Rational Choice Theory
The rapid development of rational choice theory and research has been the most dramatic change in professional political science since the 1950s. It represents a sharper break from earlier modes of inquiry than the statistically based causal modelling that has also flourished during the same period. Given the training and habits of mind that ‘rational choice’ requires, it is unlikely ever to win the allegiance of most political scientists or to have much direct impact outside the academy, but it has undeniably had a resounding impact on the more professional strata of the profession. Its root problems—the fairness of games of chance, the unpredictability of strategic interaction, the merits of different voting rules, the peculiarity of spatial competition—have more or less lengthy histories. Around 1960 the techniques that mathematicians and economists had developed to deal with these problems crystallized as a distinctive outlook and set of principles.
The principles can be summarized in three words individualism, rationalism, and formalism. Rational choice theorists seek to explain collective outcomes by individual choices, which are generally assumed to derive from fixed preferences that are basically self-regarding. Individual actors are assumed to be rational in the limited sense, roughly, of having clear goals (being able to rank the possible outcomes of their choices coherently) and being willing and able to do whatever is necessary (within given constraints) to satisfy them. But there are obviously many situations in which it is difficult to know which choices will in fact best serve one’s preferences. These situations may become clear only as the result of a ‘formal’ mathematical analysis of their elements. Consequently, it is assumed, any satisfactory explanation of what happens in these confusing situations must have the form of a mathematical model that reveals the implications of instrumental rationality.
‘Positive political theory,’ narrowly understood, refers to studies conforming to these principles. Hundreds if not thousands of investigations, by economists and sociologists as well as political scientists, could be cited to illustrate their role in contemporary political science. To ask at this point what contribution they have made to the discipline would be a bit like asking for an assessment of the contribution of probability theory or cross-tabulations. Nonetheless, the legitimacy of the approach, the value of its results, and its future prospects are now matters of heated debate.
Donald Green and Ian Shapiro, after reviewing rational choice studies of American politics up to the early 1990s, concluded that their achievements were ‘few, far between, and considerably more modest than the combination of mystique and methodological fanfare surrounding the rational choice movement would lead one to expect’ (1994: 179). Elegant as the basic theory may be, they said in effect, it adds nothing to our already large stock of knowledge about American politics and the causal processes at work in it. The cases are very rare, it seems, where an important, distinctive, and falsifiable generalization or prediction derived from the theory has not been falsified.
Stephen Walt (1999) offers a similarly harsh assessment of the contributions of game-theoretic models in international relations. After summarizing 11 exemplary studies, he concludes that ‘for malization has not led to powerful new explanations of important real-world phenomena’ and ‘recent formal work generally lacks rigorous empirical support’ (1999: 46). He ends his critique with a plea for methodological diversity, but the evidence he has assembled encourages scepticism about formal theory and the empirical research associated with it as sources of new hypotheses or well-verified findings.
Geraldo Munck (2001) makes no attempt to assess the ‘substantive contributions’ of rational choice theory in the study of comparative politics, but aims only to provide a ‘balanced assessment’ of the strengths and limitations of formal modelling as practised by comparativists. Its great strength, he suggests, is its focus on choice: despite its mathematical complexity and abstractness, it presents actors as acting, not just being pushed around by external forces. In Munck’s view, this strength is offset by some serious weaknesses, however. The expected utility model may be bad psychology; many game models have multiple equilibria, yielding no clear predictions, and in many studies ‘the rules of the game’ are treated as givens when in fact, more realistically, they are often what we most want to understand. Because of these limitations, Munck concludes, the ‘value added’ by formalization may be ‘relatively minor’ (2001: 191). In short, the reader is not encouraged to think that there are many important ‘substantive contributions’ attributable to formal modelling in the literature of comparative politics.
A deeper source of the current scepticism may be, as Munck suggests, the research by psychologists and economists on whether people generally choose ‘rationally’ in simple situations of risk and uncertainty of the kind that decision theoretic and game theoretic models are meant to represent. Do they truly want to pursue their own self-interest as it has been defined by their preference orderings and utility functions and are they capable of seeing, despite the confusing situations in which they may find themselves, what they must do in order to achieve this end? The relevant research, much of it experimental, suggests a negative answer. In other words, it seems that people tend to choose cautiously, fairly, trustingly, etc. rather than ‘rationally,’ when dealing with risk, clashes of interest, and uncertainty about how others will behave. Their ordinary conception of reasonableness evidently differs from the ‘rationality’ favoured by theorists.
One reaction to these and other criticisms has been to retreat from the demanding assumptions about instrumental rationality used in building simple models and to adopt instead more realistic assumptions as a basis for building ‘second generation models of empirically grounded, boundedly rational, and moral decision-making’ (Ostrom, 1998: 15). Such models may provide a better description of what people actually do. In principle, they could take into account the social norms, emotional reactions, moral inhibitions, limited information, limited computing ability, and cognitive crutches that seem to keep most people from being as instrumentally rational as it is easy to assume they all are. But such models, being more complex, may not yield any useful predictions or insights. The ‘folk theorem’ of game theorists—roughly that in repeated games no particular outcome can be singled out as more likely or more rational than any other—may apply, or a rule of thumb not unlike it: realism is generally bought at the price of clarity and tractability.
A contrasting reaction would be to maintain the old simplifying assumptions, but to abandon the claim that law-like explanatory generalizations or predictions can routinely be derived from them. One could even say that the simpler theories remain true, even if their predictions are always false, for they define what it is rational for individuals to do (in order to realize their postulated goals), even if nobody in fact behaves as the theories, used predictively, would require. From this perspective, the most important use of rational choice theory (or theories) would be, not as a basis for predictions, but as an aid in understanding more deeply the situations of risk, strategic interaction, and social choice in which people commonly fall into error when pursuing their own interests or advising others, and in which detached observers of their actions can also become confused. Interpreted in this way, rational choice theory remains positive and theoretical, and thus ‘scientific,’ but in quite a different way from the studies of conditioning relationships that provide the standard of comparison for critics like Green, Shapiro, Walt, and (less clearly) Munck.
Some leading rational choice theorists seem to favour this second reaction to the theory’s ‘mid-life crisis.’ Thus Kenneth Shepsle (1995) endorses the combination of ‘hard theory and soft assessment’ represented by rational choice theory, in contrast to the ‘soft (or no) theory with hard assessment’ favoured by its critics. The ‘hard theory’ offers real insight, he maintains, while ‘statistical political philosophy’ offers only unintelligible correlations. Similarly, Peter Ordeshook (1993; 1995) and Emerson Niou and Ordeshook (1999) make a distinction between science and engineering that amounts to saying that abstract models need not fit any easily observable regularities in order to be illuminating.
In fact, some of the most widely acclaimed contributions of rational choice theory have had little to do with explanation or prediction as usually understood. Thus the theory is sometimes credited with helping to revive interest in institutions, not by providing a rigorous analysis of the conditions that account for institutional differences, but rather by treating ‘institutions’ as the explanation for an otherwise puzzling fact. Simple rational choice models seem to ‘predict’ far more political instability than can be observed. Institutions can be understood as ways of constraining individual maximizing behaviour, to reduce this potential instability (Miller, 1997: 1193-8; Weingast, 1996). But how could such constraining institutions develop on the basis of individual self-interest? The recent and much discussed volume on Analytic Narratives (Bates et al., 1998) is essentially an offshoot of this ‘new institutionalism.’ Its authors aim to combine in-depth historical research with formal modelling of co-ordination problems to explain particular institutional or policy developments. Their models and parameters are chosen (and tweaked) to fit particular historical facts not to make any real predictions, or to explain any common features of all institutions, or to account systematically for differences between them (see also Bates, de Figueiredo and Weingast, 1998).
Rational choice theory thus subordinates conventional ideas about causation to the interpretation of individual intentions and the assumptions about them that underlie institutions. The explanations it provides have a fundamentally different character from those derived from theories about causal conditions. Causal analysis of the kind discussed earlier focuses on rates within classes (the Democratic share of the vote among Catholics in Detroit, and so on). Only at a limit rarely or never reached do its causal laws and statistical generalizations apply, strictly speaking, to individuals. Rational choice theory, by contrast, normally focuses on the choices made by particular individuals—not just individual persons but also other ‘individual’ actors (firms, states, parties, etc.), each of which is taken individually, so to speak, rather than as belonging to a category and being part of a comparison. The problem is to better understand the decisions these individuals make, given the situations they are in and their preferences or utility functions. They are assumed to be rational and calculating. They may in fact be caught in a vast network of physical causes and effects stretching from the distant past into the remote future—they may be cogs in a complex machine whose behaviour is determined by the past values of its variables—but this is not how rational choice theory deals with them. Whatever their causal entanglements with the past may be, they are treated as agents, not patients. They are pictured looking ahead, as it were, anticipating the consequences of their actions, not with their backs to the future, being swept along by forces beyond their control. They are assumed to be free and reasonable, at least potentially, and not just the victims of blind causation. The situations in which they find themselves may be said to cause their decisions, of course, but these determining situations have their effects, not through fixed laws or statistical generalizations linking independent and dependent variables, but through their being understood rationally (or misunderstood) as situations of choice offering each individual better or worse opportunities to pursue what he or she thinks is good.
Seen from this angle, rational choice theory represents a return to an ‘ideographic’ mode of inquiry from the currently dominant ‘nomothetic’ conception of science (Bates et al., 1998: 10). It aims at the right interpretation of individual actions, not the establishment of general laws. And yet, because of its abstractness and generality, its bold simplifications, its affinity with economics, its heavy use of mathematics, and last but not least, its unblinking acceptance of individual self-interest, not as what ought to be but as what is, it stands apart from contemporary ‘normative’ theorizing and can claim to be the most positively theoretical (or resolutely scientific) approach to the analysis of politics, exceeding in steely-eyed clarity even the most incisive statistical analyses of the hardest possible data.
The conclusion I draw is that the analysis of human behaviour can be both ‘positive’ and ‘theoretical’ without being ‘causal’ in the usual sense. Rational choice theory shows that one can focus on individuals or single cases, rather than on the statistical differences between groups or classes of individuals or cases, and that one can disregard their conditioning entanglements in order to focus on the logic of voluntary choice, without becoming explicitly normative or merely descriptive. Rational choice theory may thus help to clarify an important kind of positive theoretical inquiry that is often misunderstood today because it is thought to be either essentially normative or simply descriptive.
Clarifying the purpose or purposes of political institutions and communities is not the same as testing hypotheses about the conditions of their existence, and it may have little to do with providing simple facts about their most obvious features—their sizes, locations, budgets, the names of their officers, and the like. Nor need it have much to do with analysing the interests of individual persons. Institutions may generally serve the interests of their members, but they also shape and define those interests, and it may not be clear what the relevant interests really are. Economic institutions such as firms presumably give priority to economic goals and mainly serve individual economic interests, but it would clearly be cynical to make the same simplifying assumptions about churches and universities: no one seriously maintains that they are primarily money-making institutions. Similarly, political institutions obviously have economic functions, but they also claim to promote justice and the good life, and their various ways of understanding these goals and pursuing them raise factual and interpretive questions that invite inquiry.
Alexis de Tocqueville’s Democracy in America (1835, 1840), particularly its first volume, is an outstanding example of such inquiry. It is a description of American democracy directed to readers unfamiliar with the working of American institutions and their underlying spirit. Its presentation of American democracy may have been governed by a desire to moderate aristocratic fears of modern democracy, and its young author was certainly not shy about offering ‘normative’ advice, but the book, if it were fresh off the press today, would not be reviewed as a contribution to ‘normative political theory.’ It is far too ‘empirical’: Tocqueville’s declaration, in his introduction, that he simply wanted to make known what he had seen in America, is too close to the truth. His account of American political institutions involves assumptions about their causes and effects, including their relations to the religious beliefs of Americans, but it would be a distortion of his analysis to say that he wanted to test any general causal hypotheses. Rather, to simplify greatly, he wanted to show, in detail, the affinity between the institutions of a stable democracy and the culture or psychology—the ‘social condition’—of its citizens. American political institutions, he thought, expressed the beliefs of a people lacking high aristocratic ambitions and they encouraged those under their authority to pursue practical economic goals.
Another familiar example—a more recent classic of the same kind—is The Power Elite (1956) by C. Wright Mills. It too is fundamentally descriptive, but not simply so. It is more quantitative and social scientific in style than Tocqueville’s book, but it resembles it in subordinating causal analysis to the elucidation of collective purposes. The overall theme of the book is the transformation of American democracy since the nineteenth century. New social conditions, new institutions (national corporations, mass media, etc.), and a new role in the world have gradually given American democracy a new meaning, by contrast with the meaning it had in Tocqueville’s time.
The discussion of a community’s purposes can take a variety of forms and need not put a lot of emphasis on formal institutions, as Michael Lind (1995) shows. Indeed, it may look a lot like the analysis of causal conditions. Putnam (2000), for example, is in many ways similar to Putnam (1993), but the two books are directed to different goals. Making Democracy Work offers a causal theory based on a quasi-experimental analysis of regional differences. Bowling Alone is a more diagnostic and therapeutic investigation of contemporary American political culture. It tries to define a malaise in the way Americans relate to each other and pursue their collective purposes.
The kind of political analysis illustrated by these examples is factual or empirical and thus ‘positive,’ but neither simply ‘causal’ nor ‘rational.’ It resembles what we are doing when we try to understand the political outlook and choices of individual persons, famous or obscure. What kind of a citizen is John Doe? What does politics mean for him? What does he think it is about? How does his involvement in it fit within the story of his life? What reasons does he have for voting (assuming he does)? What considerations explain the votes he casts? Questions of this kind can also be raised about loosely organized social groupings Catholics in Detroit—as well as about organized parties and interest groups. How do they understand their situations? How do they define their identities? What are their concrete objectives? What relation do they see between the interests they wish to promote and the interests of others? How do they justify, publicly and privately, what they are doing or would like to do?
Clarifying collective purposes, particularly those of large, complex, multi-purpose political institutions and whole societies, has always been a challenge for the academic observer of politics. Only in the past century has it gradually been overshadowed by the scientific analysis of causal conditions and, more recently still, by the development of an impressive calculus of individual interests and decisions. Yet the challenge remains, as may be seen in accounts of ‘the new institutionalism’ (Hall and Taylor, 1996; Immergut, 1998) and in the literature on ‘the power of ideas’ (Berman, 1998; Blyth, 2002; Hall, 1989; Majone, 1996).
The recent literature on constructivism in international relations and comparative politics (see Adler, 1997; 2002; Checkel, 1998; Finnemore and Sikkink, 2001; Ruggie, 1998; Wendt, 1999) may provide the most revealing discussions of the problem of understanding collective purposes and relating them to the currently dominant forms of political analysis. The approach has developed in opposition to specific academic positions—neorealism and neoliberal institutionalism, for example, in international relations—and it is entangled with confusing theories about the ‘social construction’ of its objects of study. Nonetheless, it shares with a variety of recent protests against mainstream professional political science a positive analytic purpose and an insistent emphasis on the importance of shared ideas, aspirations, collective identities, and intersubjective meanings.
The common element in these examples and movements is difficult to isolate. ‘Interpretation’ and ‘thick description’ are sometimes used to define it. Charles Taylor (1971), the classic plea for interpretation in the social sciences, argued that political phenomena should be regarded as analogous to obscure texts, in need of translation or interpretive explication. As with texts, so with political phenomena: we do not understand them until we understand their meanings. Opinion polls and other surveys (e.g. Almond and Verba, 1963) may be some help, but since the relevant meanings are not just ‘subjective’ (and more or less widely shared) but also ‘intersubjective’ (and thus not normally topics for discussion or even reflection), direct answers to direct questions will often be unrevealing. The deeper meanings we seek can be brought to light only by the kind of ‘thick description’ exemplified in Clifford Geertz’s famous (1973) analysis of Balinese cockfighting.
Perhaps the best label today for what Taylor and Geertz represent is the title of this section, ‘intentional analysis.’ It avoids the unhelpful breadth of ‘interpretation,’ the novelty and obscurity of ‘thick description,’ the distracting associations of ‘hermeneutics,’ and the misleading suggestion, implicit in the old contrast between explanation and understanding (von Wright, 1971), that the clarification of intentions is not explanatory. It puts the emphasis squarely on the purposive character of individual actions and social institutions and clearly suggests the need for careful analysis, since the relevant purposes may not be obvious or easily stated. They cannot be just postulated, as they are, generally speaking, in rational choice studies, but must be investigated, for they can be complicated and obscure and may even be denied by those to whom they are rightly attributed.
The scientific status of ‘intentional analysis’ in this restricted sense is admittedly questionable. It may well seem too speculative, subjective, and impractical (too theoretical in the bad sense) as well as too descriptive (not theoretical enough in the sense of abstraction, systematic comparison, and explanation) to merit a place in political science. The evidence for its interpretations—its attributions of intentions—will generally come in the form of contestable biographical and historical narratives. Its explanatory inferences, unlike those in statistical studies or rational choice theory, will be more psychological and rhetorical than logical or mathematical. The intentions in question will be entangled with the acceptance and perhaps modification or distortion of ‘normative’ theories such as Marxism and liberalism which demand the commitment of scientific observers as well as those they observe. It will thus be much harder to maintain the standard fact-value distinction in the study of intentions (because the right description of actions will be the issue) than in the study of ‘behaviour’ (where actions are already under agreed descriptions). Finally, although an analysis of intentions may be some help to policy-makers (they may benefit from a clearer understanding of their own or others’s intentions), it is unlikely to suggest any simple formulas for influencing ‘the course of nature.’ In short, there is no denying that reasonable objections can be levelled at the idea of a social science dedicated to the elucidation of intentions (cf Winch, 1958). But there is also no denying its appeal to common sense: political institutions are purposive structures; the collective intentions that sustain them are not simply the conscious purposes of individuals at large; they are often not easily articulated, but come to light only as the result of careful investigation in a ‘positive’ spirit; and this investigation is necessarily ‘theoretical,’ not just descriptive.
Positive political theory, broadly understood, embraces most of professional political science, a discipline offering a rich array of competing approaches, methods, models, and theories. The aim of this chapter has been to clarify the current meaning of positive theory among political scientists by focusing on three main forms of it, ignoring all but a few examples of the actual research that could illustrate each type.
The statistical analysis of causal conditions and the formal modelling of rational choices are now clearly the dominant forms. The first is the direct outgrowth of the ‘behavioural revolution’ of the 1950s and 1960s. It aspires to move from the systematic collection of descriptive facts to well-grounded causal laws about political phenomena. It has shed the grandiose aspirations of earlier years and now strikes, so to speak, at targets of opportunity. It has gradually become very close in methods and spirit to applied policy analysis. Rational choice theory, by contrast, keeps alive larger theoretical ambitions. It has always scorned the vague sociological and psychological concepts, the awkward operational definitions, and the tedious statistical analyses it associates with ‘behavioural’ research. In developing its explanations, it strives to imitate modern economics, with its elegant, coherent, parsimonious, mathematical models of how people behave.
Sharp as the differences may be between statistical and rational modelling, they nonetheless share some common features. Both rely on mathematical reasoning not accessible to those without special training, and both can be understood to be contributing to causal knowledge as this is generally understood, that is, to objective knowledge of the necessary and sufficient conditions of events. In principle, the statistical analysis of independent and dependent variables goes straight for the goal, while formal modelling of the kind associated with rational choice theory approaches it by a more roundabout route. It tries to isolate and explain basic patterns of social interaction by working out the implications of individualistic assumptions about instrumental rationality.
There is little point denying that progress can be made towards the common goal by following either route. But as the possibility of future advances has become clearer, in the light of past achievements, so too have some of the difficulties to be expected and the reasonableness of some old objections. Thus, despite faster computers, larger data archives, and more powerful statistical methods, it remains true that realistic causal models of political processes often far outrun our ability to test them in any rigorous way. Similarly, findings from the experimental study of individual decision-making and advances in the theoretical analysis of strategic interaction have clarified the difficulty of deriving any solid, interesting generalizations about political behaviour from the basic premises of rational choice theory. The project of going ‘from micro-motives to macrobehaviour’ in the political realm looks more questionable now than it did a generation ago.
In the long run, the most valuable contribution of rational choice theory to political science may not be its contribution to causal analysis as commonly understood, but rather the light it can throw on the role of thought and ideas in the explanation of the behaviour of individuals. This contribution is close to the original purpose of the theory, which was to guide decision-makers in complicated situations of individual and social choice. It thus demonstrates that the ‘normative’ analysis of such situations can be closely related to their ‘positive’ description and ‘theoretical’ explanation, even when it is unrelated to the testing of any statistical generalizations. In other words, it can show that there is a causality of intentions and reasons as well as one of background conditions, and it can thus open the way for a reconsideration of a classic but now marginalized and frequently misinterpreted form of political analysis.
Alongside or beneath today’s two main contenders for the title ‘positive political theory,’ I have suggested that there is a third, ‘intentional analysis.’ And just as rational choice theory can be positive and theoretical without being causal in the standard sense, so too can the analysis (or interpretation) of the purposes actually pursued by individuals, groups, and political communities. Its rules of procedure may be less easily codified than those for statistical analysis or formal modelling; its criteria of success or failure may be less clear; its assumptions about human motivation may be far less parsimonious than the gross simplifications associated with ‘rational choice,’ but intentional analysis is nonetheless directed to answering factual questions of a theoretical character. It is not just disguised moralizing or devious prescribing—or at least no more so than the currently more reputable forms of positive theorizing. And its descriptions, like theirs, are not just collections of brute facts: they are revealing abstractions from or interpretations of the facts, showing a certain distinctive detachment from practice.
Surveying these three ways of being positive and theoretical in political science, I am struck by the power of conditional analysis to draw both rational choice theory and the study of individual cases into its understanding of the nature and purposes of theory and its relation to practice—and implicitly to exclude intentional analysis from the domain of political science. Economics is sometimes said to be a colonizing social science, using its game theoretic models and concepts like ‘social capital’ to establish its hegemony over the other social sciences (Fine and Green, 2000). But the broader and more deeply rooted imperial enterprise may be the technological way of thinking about theory, as a roundabout way of improving practical manipulation. It tends to exclude as unscientific the kind of detached observation of collective intentions that should also be called positive, political, and theoretical.