The future as arbiter of theoretical controversies: predictions, explanations and the end of the Cold War.

'The role of theory', according to John Lewis Gaddis, 'has always been not just to account for the past or to explain the present but to provide at least a preview of what is to come'.(1) Having reviewed the recent performance of international relations theorists, Gaddis also points out, correctly, that the vast majority of those theorists failed to predict the end of the Cold War. But then he argues further that star-gazers, readers of entrails and other 'pre-scientific' methods were as effective in providing foreknowledge of the fundamental transformations that took place from 1989 to 1991 as allegedly 'scientific' methods were.(2) His critique ultimately becomes an indictment not only of the inadequacies of predominant theories of international relations, but of 'scientific' methods for predicting international affairs in general. We disagree with that indictment.

Gaddis's opinions on these matters are hardly unique, nor based solely on the absence of accurate predictions regarding the end of the Cold War. Yosef Lapid referred over five years ago to the 'demise of the empiricist - positivist promise for a cumulative behavioral science', and to the 'ruins of the positivist project'.(3) In 1990, International Studies Quarterly devoted a special issue to 'Dissidence in International Studies' containing several examples of postmodern, deconstructionist, discourse analyses.(4) Donald Puchala, in an essay entitled 'Woe to the Orphans of the Scientific Revolution', defended the thesis implicit in that picturesque title by declaring that international relations theory 'does not, because it cannot in the absence of laws...invite us to deduce, and it does not permit us to predict'.(5)

Yet the scientific (or quantitative, or systematic empirical) approach to political science or the subfield of international politics survived - and even flourished through - earlier celebrations of its demise. David Easton, for example, praised the 'post-behavioral revolution' in political science long ago.(6) Our central claim here is that current reports of the death of the scientific study of international politics are also exaggerated. In support of that claim, we will discuss several issues regarding the use of predictions in the evaluation of different, competing approaches to international politics. We then evaluate the ability of the scientific approach to international politics to produce accurate predictions about political events in general, and about the end of the Cold War in particular. We conclude that the thesis regarding the demise of the systematic empirical study of international politics advanced by Gaddis and others ignores a substantial body of work which we believe justifies a more positive judgement.

PREDICTIONS, EXPLANATIONS AND USEFUL THEORY

Gaddis asserts that theories should highlight patterns from the past in a way that makes them useful guides to the future.(7) Hans Morgenthau argued that 'realism' would allow analysts not only to foresee but to influence the future.(8) No less an authority on science than Stephen Hawking declares that 'a good theory ... must make definitive predictions about the results of future observations'.(9) Similarly, Robert Keohane in his essay on 'Theory of World Politics' emphasizes that foreknowledge is one of the most important products of good theory.(10)

But J. David Singer, a founder of the quantitative approach to international politics, argues that prediction demands less of a theory than does explanation, or even description. He points out, for example, that it is relatively simple to predict that pressure on the accelerator of a car will increase its speed. Explaining why that is the case is a more demanding task, in his view.(11) Joining Singer in this sceptical attitude about the utility of accurate predictions as an 'acid test' of the validity and quality of theories are rather strange bedfellows: an array of 'anti-' or 'post-positivists', and more currently of postmodernists who disagree with the idea that hypotheses can be tested by comparing their implications with experience. In the postpositivist view, there is no way of describing or observing the 'real world' in a way that is independent of the theory that produced the hypothesis. This means that no one can 'objectively' observe the correspondence between theory-based predictions and happenings in the 'real world'. 'There are no 'brute' facts - no facts prior to interpretation ... Facts are always theory-dependent'.(12) Postmodernists 'consider both [causality and prediction] uninteresting ... They argue [that] the requirements of temporal priority and independent, external reality assumed by these concepts are dubious'.(13)

Some postmodernists also feel that any correspondence between theory-based predictions and 'real world' events will be doubly misleading, not only because the perceived correspondence will be produced spuriously by theory-driven perceptions of the 'facts' in question, but also because any correspondence that is perceived will also be in part the result of a self-fulfilling prophecy. That is, being aware of the theory, people will behave in the way it predicts because they believe in its validity, or because they think that the common expectations produced by it in the culture in which they operate make it 'rational' to behave as the theory predicts.(14) Related to this idea is the postmodernist belief that theoretical predictions constitute a form of political oppression. In this view, communities of people ought to be allowed to supply their own meaning and truth for themselves. Any attempt to generalize across communities involves the destruction of differences. Such generalizations are in effect 'hegemonic power plays'.(15) In the opinion of Richard Ashley, 'Any knowledgeable practice that participates in the inscription of sovereign voice and the narrative structure of history - and the conduct of social theory is certainly one such practice - is an arbitrary practice of power'.(16)

EVALUATING THEORIES: THE CASE FOR PREDICTIVE ACCURACY AS A CRITERION

In contrast, A. F. K. Organski and Samuel Eldersveld, for example, find 'troubling' the persistent view that the capacity to predict is of little consequence because explanation is the real commitment of the social sciences.(17) One of the origins of such discomfort is the influential argument of Carl Hempel that 'the logical structure of a scientific prediction is the same as that of a scientific explanation'.(18) This argument implies that explanations and predictions should both be defined as contingent statements, and contrasted with 'prophecies', or assertions about the future prefaced by no conditions or contingencies.(19) It is true, as Singer has argued, that 'any informed layman' can predict that if pressure is applied to an accelerator of a car, it will increase its speed.(20) But the prediction will be wrong if the car has run out of fuel. The contingency 'if the tank contains petrol', or any other contingency that would improve the explanation might also usefully be considered part of a more complete or accurate prediction.

This is not to say that we accept unconditionally Hempel's argument that explanations are identical in structure to predictions, nor are we advocating what some would term a 'naive falsificationist' view that scientists are analogous to bookkeepers who can mechanically tally up the extent to which the predictions based on a theory are supported by events in the real world. We acknowledge, for example, that scientists can be and often are influenced by predominant 'paradigms' or 'research programmes' in ways that make them quite different from accountants adding and subtracting columns of numbers.(21) We are also aware of, and in partial agreement with, scientific 'realists' who contend: that explanations focus on causal mechanisms and processes; that many intuitive notions about explanations and causality do not fit into Hempel's arguments regarding the symmetry between explanations and predictions; that many successful explanatory schemes do not necessarily provide a basis for accurate predictions; and that in general explanations cannot be equated with predictions, or vice versa.(22) But the Hempelian and the 'realist' conceptions of explanation (and prediction) are not in irreconcilable competition with each other. 'The two explanatory modes are complementary; a full science will recognize both'.(23)

While we acknowledge, then, that explanation may be in general a more demanding or important task for theory (and theoreticians) than prediction, we are also inclined to argue that in one important respect, at least, predictive accuracy is a more stringent criterion. Explanations may be - but predictions cannot be - modified, consciously or subconsciously, in order to accommodate the events upon which they focus, since the outcomes to be accounted for by predictions are unknown. This makes the future an important, even irreplaceable, arbiter between contrasting claims based on competing theoretical or epistemological approaches. In other words, we have no quarrel with those who insist that 'understanding', not prediction is the main purpose of science. But 'prediction can be a purpose too ... It is the test of a theory, whatever the purpose'.(24)

We adopt such an argument even though we agree with postpositivists, postmodernists, and hermeneutically inclined analysts that there are no 'facts' out there in the real world to be observed independently of theoretically derived expectations and inclinations. Since human beings cannot be totally objective, even the most seemingly innocuous descriptive observations about international politics are 'theory-laden'. Consider the assertion that 'Germany initiated the Second World War in 1939 by attacking Poland'. It involves anthropomorphizing the states in question ('Germany' and 'Poland') in a way which is not only theoretically based, but objectionable to some scholars of international politics. The phrase 'Second World War' is ethnocentric. It might, more objectively, be referred to as the second phase of a European Civil War.(25) To designate the point in time as '1939' is equally ethnocentric, accepting the Gregorian calendar as opposed to the Chinese or Hebraic methods of assigning numbers to the passing years. Potentially even more substantial issues raised by this statement include whether Germany really was responsible for initiating the Second World War, or when it really began.

But to go from a valid assertion that observations of facts about the 'real world' are inevitably theory-laden to the conclusion that one cannot usefully compare derivations from theories to the results of empirical observations is to press a valid point to an extreme which is unnecessary. 'Difficult cases make bad law'. Analogously, it is a mistake to conclude from the impossibility of making 'neutral' or 'objective' observations of the real world that it will be impossible to achieve any meaningful degree of consensus about whether predictions derived from a theory are accurate or valid. In short, we acknowledge that observations are inevitably theory-laden; they are not, however, theory-determined to the degree that comparing observations to theoretical expectations is a pointless exercise.(26)

Admittedly, it is possible that a majority of observers will form a theoretically based consensus around ideas that only appear to be validated, because the social actors being observed are infused with the same theoretical notions and behave accordingly. But especially if one considers the number of null findings that never see the light of published day, then the proportion of empirical analyses that fail to support the hypothesis in question suggests that most of them are not self-fulfilling prophecies. In addition, many analyses address the past, sometimes more than two millennia ago. Since the theories in most cases did not exist during the times on which such analyses focus, there is no danger that they will be confirmed in a self-fulfilling way. Similarly, the geographic scope addressed by predictions in the field of international politics, at least, militates against the self-fulfilling prophecy. Such points are even more compelling in the light of the significant number of competing, diametrically opposed propositions in the field of international politics, even though their adherents share epistemological commitments to systematic empiricism.

In response to the fear that accurate predictions (or the theories on which they are based) constitute a form of oppression, we would not deny that those who attempt to predict political events typically have an interest in having an impact on them. In fact, scientifically oriented analysts will often proclaim proudly that their work is 'policy-relevant'. Nor is it possible to deny that the knowledge producing accurate forecasts of political processes or events can fall into the hands of people who will use it for nefarious purposes. But to eschew the pursuit of scientific knowledge because of its potential for exploitation for undesirable ends is a step towards ensuring that fate for such knowledge. In other words, if 'good' people (by definition) avoid the generation of scientific knowledge on the grounds that it can serve undesirable ends, then only 'bad' people will produce such knowledge. 'Although science may or may not require prediction, we do. It is a large part of what we look to science for. It is what we need to ameliorate the human condition or at any rate to prevent further deterioration'.(27)

Fortunately, too, science as a human endeavour is opposed in principle to the imposition of ideas by authority, or by the power inherent in formal authoritative roles. An essential character of scientific evidence is that it be reproducible that anybody, but especially critics or opponents of ideas espoused by scientists, should be able to reproduce the results or evidence in support of those ideas on her or his own, independently. Admittedly this is an ideal often not attained in the 'real world'. But even approximations of the idea provide some protection against the use of scientific knowledge for oppressive purposes.

PREDICTING THE POLITICAL FUTURE

The assertion by John Gaddis that the field of international politics has failed to provide a basis for making accurate predictions about political events (such as the end of the Cold War) is based on a thorough review of theoretical and empirical work in the field. But he does not mention a set of related streams of research and theory that justifies, we believe, a more optimistic evaluation of the field's ability to deliver accurate predictions.(28) The streams of research to which we refer are, specifically: a rational choice approach to political forecasting; the wisdom and insights of traditionally trained country and regional specialists upon which that approach relies; and the developing body of work devoted to an elaboration and evaluation of the democratic peace proposition.

The origins of the political forecasting model based on rational choice theory can be traced to The War Trap by Bruce Bueno de Mesquita.(29) The theory introduced there was refined in 1985,(30) and served in turn as the basis for a model designed to produce forecasts of policy decisions and political outcomes in a wide variety of political settings. 'The expanded domain of the theory required no modification of the formal …