20th WCP: The Duhem-Quine Thesis in Economics: A Reinterpretation

Introduction: The Duhem-Quine Thesis in Economic Methodology

The Duhem-Quine (hereafter the D-Q thesis) thesis occupies a peculiar position in the writings of contemporary economic methodologists. For some it is credited with generating major developments in philosophy of science. Blaug, for instance, within the context of his discussion of Popper's methodological falsificationism, argues that Popper was "not only aware of the Duhem-Quine thesis but indeed the whole of his methodology is conceived to deal with it" (Blaug 1980:18). Hands, in a discussion of Lakatos's methodology of scientific research programmes, argues that it was largely the existence of the "Duhemian Problem" which led Lakatos in the first instance to abandon falsificationism and to develop the methodology of scientific research programmes (Hands 1979:295) Arising from the perceived centrality of the D-Q thesis one could reasonably anticipate an extensive literature on its implications for economic methodology. This anticipation would be further reinforced by the fact that the social sciences in general and economics in particular are clearly subject to the Duhemian stricture of the absence of crucial experiments in the domain of theory appraisal.

Within the economic methodology literature, however, the D-Q thesis has not been the subject of the extended analysis that one might have expected. Seminal contributions by Cross (1982, 1984) attempted to address the implications of the D-Q thesis in the context of the testing of a specific economic hypothesis, i.e. the stability of the demand for money function. Cross invoked Lakatos' methodology of scientific research programmes and argued that a modified version of it provided an adequate response to the epistemological problems identified by the D-Q thesis for economic methodology. Though Cross's response was challenged by Heijdra and Lowenberg (1986), for instance, it represented the first explicit attempt to examine the implications of the D-Q thesis in the economic methodology literature and is the most cited contribution on the topic to date. A much more recent contribution by Sawyer, Beed and Sankey (1997) follows the spirit of Cross's contribution by selecting four specific hypotheses in economics and by examining the implications of the D-Q thesis for the testing of these hypotheses. Their work provides an interesting attempt to generate a taxonomy of auxiliary hypotheses and to test the robustness of variation in these different types of auxiliary hypotheses. They conclude that only with respect to quite specific types of auxiliary hypotheses can the effects of the D-Q thesis be lessened and they maintain that the challenge for future work is "to diminish the effects of the Duhem-Quine thesis by constructing theses robust to many types of auxiliary hypotheses "(Sawyer, Beed Sankey 1997:21).

What these contributions have in common is a clear focus on the implications of the D-Q thesis for the testing of individual hypotheses in economics — in the case of Cross a single hypothesis is examined, while Sawyer et alia select four particular hypotheses. More recently a new development has emerged in the literature which invokes the D-Q thesis to highlight the difficulties and ultimately the impossibility of testing at a more aggregate or holistic level of analysis, namely schools of thought or what Kuhnians call paradigms or major trends within particular paradigms. This shift in focus away from individual hypotheses to extensive complexes of sentences, such as neoclassical economics or trends in post-Keynesian economics, is evident in the recent work of Hausman (l992) and Boitani and Salanti (1994) respectively.

In an engaging discussion of the role of theory in the context of their analysis of different strands of Post-Keynesian economics, including new Keynesian economics, Boitani and Salanti (1994) concluded that just two possibilities presented themselves. Accepting some form of methodological and theoretical pluralism, the authors argue that one either confines oneself to "internal" criticism, i.e. to appraising the sole inner consistency of each theoretical perspective with its own methodological justification (Salanti 1989 a, b) or alternatively, if one continues to maintain that economists should not surrender to the view that economic theories are incommensurable, then one must be prepared to embark on protracted methodological discussions on how economic theories or models are to be empirically appraised. In connection with the latter possibility the authors conclude "one is likely to go through the Duhem-Quine thesis on the impossibility of testing single hypotheses within complex scientific theories" and that such "an impossibility undercuts all forms of testing, particularly in a discipline like economics, the subject matter of which can hardly be submitted to experiment and is continually changing in historical time" (Boitani and Salanti: 1994: 148 italics ours). Likewise Hausman (1992), though motivated by different methodological considerations to that of Boitani and Salanti, also invokes the D-Q thesis in the context of his analysis of neoclassical microeconomics. Hausman also adopts and extremely pessimistic interpretation of the D-Q thesis. If, as Hausman argues, "the Quine-Duhem problem is posed as a purely logical difficulty, then it may not be in practice very serious". But, he continues, "if one is unable to place much confidence in the other premises needed to derive a prediction P from an hypothesis H, then there is a serious practical problem. Indeed it becomes almost impossible to learn from experience. This is the situation in economics" (Hausman 1992: 307, italics ours). In this sense according to Hausman economics "is cursed" (Hausman 1992:226)

In the following section we will outline how causal holism, first elaborated in Boylan and O'Gorman (1995), re-engages the D-Q thesis. In this connection causal holists relocate the D-Q thesis in its original context, i.e. Quine's critique of logical positivism, and we show how this relocation and reengagement liberates economic methodology from the undue pessimism of the Hausman-Boitani-Salanti interpretation.

Causal Holism and Economic Methodology: Quine Revisited

According to causal holism Quine's philosophy of science is neither fully nor properly appreciated in recent debates in methodology of economics, especially in realist analyses ranging from Lawson to Mäki (Boylan and O'Gorman 1995). The spirit of causal holists's approach to Quine is summed up in Keynes's famous tribute to Marshall (with the obvious modification i.e. the substitution of the term "economic" by the term philosophical): "...those individuals who are endowed with a special genius for the subject and have powerful economic intuition will often be more right in their conclusions and implicit presumptions than in their explanation and explicit statements. That is to say their intuitions will be in advance of their analysis and terminology" (Keynes 1924:335). In other words Quine's works are rich in gold but they are not all gold.

For the purposes of a causal holist interpretation of Quine in the context of economic methodology we will limit our discussion to three issues: economic theory, the testing of individual economic hypotheses and finally the testing of an economic theory or a major trend in an economic "school". As we have already noted, the standard literature is focused on the impact of the D-Q thesis on the testing of either individual economic hypotheses or economic theories. As Cross pointed out, the D-Q thesis, under the shadows of Popper and Grünbaum, entered mainstream methodology of economics in connection with the issue of the testing of individual economic hypotheses (as distinct from an economic theory or framework such as neoclassical economics). Causal holism initially shifts the focus away from this specific issue to a more general issue which historically was integral to the original context of Quine's famous thesis, namely his critique of logical positivism and his construction of an "empiricism without the dogmas" (Quine 1953).

Quine's critique of the positivist approach to science focused on its atomistic approach to scientific language and its prioritization of a purely descriptive, theory-neutral language. In opposition to this, Quine developed a holistic view of scientific language in which scientific descriptions are theory-laden. This Quinean holistic approach to scientific language is a major influence on the causal holist approach to economic methodology. As it developed in historical time, economic language is not independent of economic theory. A purely descriptive language, totally uncontaminated by theory, is not available to economists in their descriptive endeavours. The individual descriptive statements of economists are inextricably embedded in a holistic web of interrelated sentences which in turn are holistically connected to theoretical sentences. The descriptive language of economics is theory-laden.

Unlike Cross, causal holists do not start with the D-Q thesis at the level of testing. Rather they distinguish sharply between the D-Q thesis at the level of meaning and at the level of testing. More precisely, causal holism distinguishes between a Quinean holism at the level of meaning and any possible holism at the level of testing. Its primary commitment is to holism at the level of the meaning of economic statements. The meaning of an economic descriptive statement does not reside in the statement the way a nut is contained in its shell. To grasp the meaning of an economic descriptive statement, the statement must be located in the appropriate web of other sentences including some theoretical ones.

This holistic approach to economic language forces causal holism to challenge an influential conception of economic explanation, based on the assumption that the aim of economics is to accurately describe the economic world in a theory-neutral language and, having accomplished this descriptive task, to construct the correct or approximately correct economic explanation of these economic facts by recourse to economic theory. For causal holists this separation of description from theory does not stand up to critical scrutiny. Contrary to the above conception where economic theory is the hand maiden of economic explanation, for causal holism economic theory is the hand maiden of accurate description. In a causal holist framework, a central aim of economics is the construction of models which furnish accurate descriptions of the functioning of economic systems and agents in historical time and economic theory is an indispensable source of the economic concepts and language used in the construction of these models. There is no such thing as a theory-neutral, atomistic language adequate to our economic descriptions. Thus causal holists defend Marshall's famous remark that economics "is an engine for the discovery of concrete truth" in the sense that without a sophisticated "engine" of language we will fail to articulate accurately the concrete truth about real economics. In short causal holism critically engages the received wisdom of the role of theory in economics and, in light of Quine's holism, centrally connects economic theory to description rather than explanation.

We now turn our attention to the issue of testing. A basic question for causal holism is the following. What price must be paid at the level of testing for a Quinean holism at the level of meaning? In this connection, contrary to a standard interpretation of the D-Q thesis which distinguishes between a strong and a weak thesis (Heijdra and Lowenberg 1980:176), causal holists introduce a threefold distinction between the weak, the stronger and the strongest D-Q thesis (Boylan and O' Gorman 1995: 78-82). The weak D-Q thesis highlights Quine's new empiricism in which holistic testing is indispensable to any science. There is no science without the bar of experience. Quine shares this thesis with logical positivism. What is wrong with logical positivism is that it gives an erroneous account of testing arising from its atomism. The weak D-Q thesis applies to either individual statements or a theoretical system (or subsystem). Individual statements come before the bar of experience holistically: they are not tested in isolation. In the case of a theoretical system, such as neoclassical economics, it too comes before the bar of experience. The weak thesis is silent on how this testing is to be accomplished. However, it insists on its accomplishment.

The stronger version of the D-Q thesis is taken from Hesse: "no descriptive statements can be individually falsified by evidence, whatever the evidence may be, since adjustments in the rest of the system can always be devised to prevent falsification" (Hesse 1970: 195, italics ours). The weak thesis makes no reference to the issue of adjusting the system for the purposes of preventing the falsification of some preferred descriptive statement. The stronger thesis raises this issue and focuses on the above logical possibility. While causal holists fully acknowledge this possibility it queries whether it is always wise or prudent to exercise this possibility. Causal holists maintain that circumstances can arise where one would be totally lacking in judgement to retain a specific hypothesis by readjusting the rest of the system. Thus in economic methodology it welcomes the research of Sawyer et alia (1997) which, as we already noted, furnishes a taxonomy of auxiliary hypotheses and tests for their relative robustness. Such classifications and tests for robustness constitute a range of tools for rational choice. However, such choices are not fool-proof: wise experienced economists will be guided by these tools in the methodological knowledge that creative economists will invent novel tests which may well force them to either abandon or reshape existing tools. In short, while causal holists accept the stronger version of the D-Q thesis as a logical possibility, they recognize that specific circumstances could make its actualization at best unwise or ill judged.

The strongest version of the D-Q thesis is summed up in Quine's famous dictum "any statement can be held true come what may, if we make drastic enough adjustments elsewhere in the system" (Quine 1953: 43). The strongest version is quite radical and has been interpreted in a relativist way (Laudan 1990). Unlike the stronger thesis, it applies to every kind of statement, descriptive, theoretical, lawlike, mathematical, a priori, logical and so on. Also unlike the stronger, it goes beyond the issue of falsifiability to the issue of truth. It could be argued that, since the weak and the stronger D-Q theses are implied by the strongest, Quine himself endorses all three. Without prejudice to Quinean scholarship, causal holists reject the strongest thesis while retaining the weak and the stronger theses outlined above.

In view of its acceptance of the weak D-Q thesis, causal holism is obliged to give an account of how an economic theoretical system or framework or subsystem, as distinct from an individual hypothesis, comes before the bar of experience. Perhaps Hausman (1992) and Boitani and Salanti (1994) are correct in their pessimism vis-à-vis testing in economics? In Hausman's colourful terminology neoclassical economics is "blessed" in its mathematical rigour and other internal virtues but is "cursed" in that recourse to empirical testing will always remain inconclusive (Hausman 1992:276). Under the influence of Kaldor's methodological writings, causal holists do not share Hausman's pessimism. Kaldor was an intriguing economic figure in that in his early economic research he made an invaluable contribution to neoclassical economics. In his middle period, however, he turned to Keynesian economics while in his later writings he became an uncompromising critic of general equilibrium theory, the jewel in the crown of the neoclassical economics he espoused in his early period.

In developing his insightful critique of general equilibrium theory, Kaldor took a broader, more holistic approach than many of his other Cambridge colleagues who shared his dissatisfaction with orthodox economics. In Kaldor's view there is not "a single overwhelming objection to orthodox economic theory: there are a number of different points that are distinct though interrelated" (Kaldor 1975:347-8). Indeed he explicitly considered some of his Cambridge colleagues who also shared his opposition to orthodox economics as "monists" in that they held that the exposure of the flaws in the orthodox theory of marginal productivity was "alone sufficient to pull the rug from under the neoclassical value theory" (ibid.). According to Kaldor this monist approach is too narrow: there are "other things to object to that in some ways are even more misleading than the application of marginal productivity theory....which has been the main subject of discussion". (ibid.).

According to causal holism Kaldor's "non-monist critique" offers a clear paradigm of holistic empirical testing in line with the demand of the weak D-Q which requires a theoretical system to openly face the risk of being rejected by the bar of experience. Basically this consists of two steps. Firstly the construction of a wide range of economic models, based on the conceptual resources of equilibrium economics, purporting to describe real economic sectors as these develop in historical time. Secondly demonstrating, by recourse to a variety of means, how these descriptions are empirically inadequate (cf. Boylan and O'Gorman 1995, 1997 for more details of this holistic testing). In this Kaldorian approach there is nothing akin to a crucial experiment - it is failure over a wide range of models which legitimates the rejection.

Of course in this Kaldorian holistic approach it is possible to retain the core of orthodox economics by making radical adjustments elsewhere in the system. However, for the causal holist, a Kaldorian type weight of negative evidence across a wide range of models furnished by orthodox economics would imply that the retention of the core on the grounds of it being internally blessed is at best ill-advised and at worst ill-considered. Contrary to Hausman orthodox economics is not destined by any hidden hand of any god to remain forever the chosen one. If one's methodology is empirical as in causal holism, it can and must fact the tribunal of experience.