The Keynesian Method, Complexity, and the Training of Economists

Abstract

This paper argues that there are serious problems both with the grand synthesis IS/LM macroeconomics and with the modern DSGE macroeconomics. It argues that the reasons for the problems are not with the models per se but with the way the models are used. Specifically, it argues that both approaches have deviated from the Keynesian method, which in the author’s mind, should have been the most important legacy of Keynesian economics. The paper provides a brief history and explanation of how the profession evolved from Keynes’ writings to the modern DSGE approach. It argues that the problem was that academic incentives drove the profession away from the Keynesian method and instead directing economists to become more concerned about publication than about understanding. The paper concludes by explaining why the loss of the Keynesian method is of concern.

Keynes is dead; dynamic programming; Keynes is still dead. That’s the way Stanford graduate economics students recently summed up what they had learned in their core graduate macroeconomics course (Colander 2007: 152). Graduate students at other top U.S. programs concurred, and in my recent interviews with them the strong feeling among students was that the core graduate macro course provided them little in the way of macro policy thinking and that the course had nothing to do with Keynes. It was a course in which, by design, students learn dynamic stochastic programming.¹ A comment of an M.I.T. student was representative of students’ view of what they learned about policy in macro. He stated: “Monetary and fiscal policy are not abstract enough to be a question that would be answered in a macro course” (Colander 2007: 169).

As Robert Solow (2007) points out, the situation is a far cry from the core macroeconomics courses of the grand macro synthesis period of the 1950s–1970s period when students studied “Keynesian synthesis” models and macro policy in their core courses. In the synthesis period students learned variations of IS/LM models, and how that IS/LM type of reasoning underlay both large econometric macro models and macro policy. Compared to modern DSGE models, the synthesis models were technically simple, and macro graduate student training of the time was not highly technical. It involved a blend of institutional, historical, and policy training. In this grand neoKeynesian/neoclassical synthesis, macro theory, empirical work, and policy were entwined in a superficially connected, but ultimately unsatisfying, set of models that built differences of policy position on slim reeds such as Pigou effects and wage and price rigidity assumptions. The models could be adjusted to “explain” just about any observation, which meant that, more often than not, researchers’ judgments determined the results of the model. By that I mean that it was a synthesis in which one could predict the policy implications of models based on who was doing the modeling. It was hardly a situation that inspired confidence in the usefulness of the models.

As should be clear from the above description I am no fan of the grand synthesis macroeconomics. But I am also no fan of modern DSGE macro as a basis for policy analysis. Robert Solow nicely captured my view of current macro theory when he wrote that modern macro is best seen as a “rhetorical swindle” that the “macro community has perpetrated on itself, and its students” (Solow 2008: 235).

In this paper I (1) explain why I am not a fan of either grand synthesis macro or modern DSGE macro and how both have deviated from the Keynesian method, which I believe should have been the most important legacy of Keynesian economics; (2) provide my explanation of how the profession moved from Keynes’ writings to the modern DSGE approach; and (3) explain why I believe the loss of the Keynesian method is of concern.

Paper Prepared for the First Bi-Annual Symposium, The Thomas Guggenheim Program in the History of Economic Thought, “Perspectives on Keynesian Economics”, 14–15 July 2009, Ben-Gurion University.

Comments on Colander’s “The Keynesian Method, Complexity, and the Training of Economists”

Amos Witztum

Colander’s paper is a fascinating attempt in understanding how did we get to where we are in Macroeconomics given the Keynesian starting position. The focus of the paper, however, is on methods rather than substance. It is an unusual approach but appears to be quite rich in content and controversy. The paper deals with three main issues: the methodological significance of Keynes’s revolution, the process which led to the current dominance of the Dynamic Stochastic General Equilibrium (DSGE) models, and, the question of theory–policy relationship. The general thesis of the paper is that the Keynesian method was more like an ad-hoc approach to economic studies which accepted the existence of many models and combined their choice and use with intuition and some knowledge of the real world. Behind it, however, there is a complex system of the macro economy which combines many aspects of social life. But the complexities of social life were reduced by economists to that which is easier to handle: mathematically sophisticated models. The reason why these models developed is mainly due to the fact that while in earlier years economists were trained by people with broad interests, since the 1930s economists have been trained by increasingly narrow minded economists. Coupled with the ease of establishing one’s self academically through mathematical complexities rather than real struggle with the social structures, we have landed in this new world of DSGE models that are as good (or, actually, bad) in explaining things as their predecessors.

I have great sympathy (and agreement) both with the claim that the current state of macroeconomics is lamentable and that the search for ever sophisticated mathematical models is more the outcome of the sociology of the subject than a service of the discipline’s intellectual needs. But one can reach similar conclusions without adopting Colander’s take on how economics should be done either as an interpretation of the Keynesian method or as a general methodological claim.

Keynes’s Method

According to Colander, Keynes would never move straight from a model to a policy recommendation. Nor would he consider only one model. Instead, there will be some kind of a Meta theory at the back of one’s mind. When addressing policy issues, one will then be engaged in the art of selecting a model which could be useful for the question at hand. This, as Colander argues, would be consistent with the views of Keynes’s father as well as Robbins’s conception of economic science.

I completely agree that in terms of Robbins’s conception of economics, the only way to fit Keynes’s work into Robbins’s definitions of the subject would be to treat it as a policy document. Moreover, my own feelings are that while Keynes did not offer much more than an ad-hoc theory, it was not of the kind which Colander describes as a Keynesian method. However, there is plenty of textual evidence to suggest that Keynes did not consider his work in this narrow way. Quite to the contrary, the mere title of the book – The General Theory – does not seem to support the idea of a modest contribution to resolve an ad-hoc problem. In fact, it seems to hint at much grander objective akin, one could say, to Einstein’s contribution to Physics¹⁵:

“The classical theorists resemble Euclidean geometers in a non-Euclidean world who, discovering that in experience straight lines apparently parallel often meet, rebuke the lines for not keeping straight – as the only remedy for the unfortunate collisions which are occurring. Yet, in truth, there is no remedy except to throw over the axiom of parallels and to work out a non-Euclidean geometry. Something similar is required in economics.” (GT, 16)

Notwithstanding the ambiguity which is revealed in this quote with regard to Keynes’s views on classical economics (is it like Galilean relativity or Special relativity in its relationship with the General theory?), it does not sound anything like the method which Colander attributes to Keynes. Moreover, as I say in my own contribution to this volume, there can be little doubt about Keynes’s own sense of revolution. “I find myself” he wrote to G B Shaw, “to be writing a book on economic theory, which will largely revolutionise…the way the world thinks about economic problems” (Skidelsky 1992: 520). I believe that it would require a great leap of faith to believe that addressing a specific problem in an ad-hoc manner – which is what Colander calls the Keynesian Method – constitutes something which would revolutionise the way we think about economics. After all, it is not really the case that real classical economists like Smith, Ricardo and Mill did not write on policy issues without being completely bound by a restrictive model deduced from first principle.

So it seems that Keynes was clearly aiming at something much greater than the mere solution to the policy issue of dealing with unemployment. I believe that this is enough to disqualify him from following the method which Colander attributes to him.

Still, if we read Keynes as he really is then there may be something in Colander’s claim. The fact that Keynes decided not to touch upon the question of economic organisation may fit Colander’s argument that he “chose” a model from an existing arsenal and adjusted it so that it would fit the problem he was dealing with. Nevertheless, the discrepancy between the man’s intentions and achievements makes it difficult for me to accept that this which he did not really intend should become the defining factor of his legacy.

I must confess that what Colander calls Keynes’s method sounds very similar to his own earlier idea about ad-hoc economic analysis. In his excellent paper from 2000 Colander argues that the ad-hoc approach where we address individual problems at a time without a direct reference to any paradigm would be the natural development once people realise the futility of what he calls “structural simplifications” which represent formal modelling of the economy. In fact, he even suggested a sequence that would ultimately lead to a “Keynesian method.” At first, there will be the stage which Solow called “loose-fitting positivism” where the ad-hoc work is based on models which are consistent with the main paradigm but this will give way to “pragmatism” where “the models actually being used for policy purpose were diverging from the underlying formal general equilibrium models at their core” (Colander 2000: 127). In other words, my feeling is that the issue is really Colander’s own frustration with current DSGE models rather than a cry over the betrayal of Keynes.

Complexity and the Segregation of Theory from Policy

I fully agree with Colander’s analysis of how economists moved from sensible writings to increasingly more complex mathematical models with increasingly more tenuous relationship with their subject matter. I agree with his description of the increased poverty of economics teaching and the senseless rise of quantitative based stature within the discipline.

However, I am not entirely sure that I have been persuaded by his proposed disjunction between theory and policy. On the one hand, Colander accepts that there is a need for a Meta theory and he concedes that this must have also been in Keynes’s mind. He rightly deplores the decline of this pursuit into the mundane activity of excessive formalism but he does not offer any insight into how it should be done. Instead, he seems to be more concerned with separating the policy people from the theorists. This, he would like to see happening in both training and in research.

To some degree, Colander argues, the IS/LM synthesis was a good model as it has its legs, so to speak, in both theory and practice. DSGE, on the other hand, seems to him much less institutive and not much better in providing predictions. However, he then goes on to say that “the DSGE model is a much better scientific model than is the IS/LM model” (p. 14). This sounds to me very strange. How can something which is totally useless be of any scientific value? If the model does not offer a reasonable explanation of the world nor capable of predicting it, how can it be a much better model? Is it just because of its mathematical sophistication? All of a sudden, all those things – like senseless and excessive formalism – about which Colander rightly complains – become the essence of the science of economics? I am not at all surprised that if this is his vision of theory, that policy should have nothing to do with it.

But the real question is whether economic science (theory) should develop in such a way as to become incomprehensible for the purpose of policy making. Because Colander says little about how the science of economics should progress, he wants the people who do policy to be taught about the world more than they are taught about theory.

Unlike Colander, I believe that the real issue is to address the question of how can a theory which is aimed at explaining the economic world be so useless for the purpose of policy making. In other words, the question is not really how to separate policy from theory but rather, how to do theory. Colander only touches upon this in his discussion of complexity. Basically, he claims here (and in his 2000 paper) that the complexity of the world calls for complexity in science. Thus, there are phenomena which are susceptible to traditional economic modelling (“linear dynamics and unique analytical solutions” (Colander 2000: 128)) and those which require a “replicative process” where research begins with a variety of organising principles and dynamics.

I am not entirely sure that I recognise these different phenomena about which Colander writes. It seems to me that he is referring here to the question of the relationship between economics and the other social sciences. I suppose that what he means here is that the phenomena which lend themselves to traditional deductive reasoning are those which can be identified as economics phenomena and the others, the more complex ones, are social phenomena. Personally, I am not sure that I would go along with such a distinction. In some ways, this seems to raise a similar question to the one raised during the famous “Methodenstreit”: can we separate the economic aspects of social phenomena?

If the answer is no then there is no purpose in pursuing the analytical advancement of economics as a separate discipline. The recipe of “replicative process” should be the only rule and by implications, economics should never be taught as an independent disciplines. While there may be many people who believe this to be true, I am not quite sure that it is. There can be little doubt that economic phenomena are social phenomena. There is also no doubt that one must take into account the other aspects of society when investigating such phenomena. But this does not mean that one cannot conduct a ceteris paribus examination provided that one also has a good understanding of those things which are held fixed and which, over time, must never be kept in the same position. In fact, classical economists like Smith or J S Mill had done exactly this. They created a coherent system of economic analysis which is closely tied together with other aspects of social life. Whether or not their systems lend themselves to a mathematical formulation is neither here nor there in terms of the development of our understanding of how social organisation works. The fact is that they have offered a genuinely complex simplified structure that does not require the separation of policy from theory in such a harsh way as proposed by Colander.

In this respect, Keynes – very much like today’s DSGE models – are guilty of the same thing. They have transferred the economic problem from being a problem of social organisation (the subject of complex science) to being an ad hoc problem of how to determine and control aggregate values. At the time of Keynes this was a betrayal of the agenda of the real classical economists – unless you see his work as a policy document – and in this respect, the DSGE models are true to Keynes’s tradition.

Conclusion

Colander’s paper makes the case for a separation of theory from policy. In this respect, he follows Robbins even though he does not make the demand that theory should also be value free. It is based on his justified (and well documented) frustration with the over formalistic drive which left economic theory like the DSGE divorced from intuition and real life experience. He has also highlights and deplored the sociology of the disciplines which allows for such developments. For this, he should be congratulated.

Colander has obviously thought a lot about these questions and developed his own vision of how things should progress. These include both a division of labour between theory and policy and, a division of labour between “structural simplifications” and “replicative processes.” In the paper in this volume he attributes a similar way of thinking to Keynes. While I have not been persuaded that he is right about Keynes, the vision is still relevant. Here, however, I felt that I have to differ. While the integration of economics into the social sciences may indeed make Colander’s vision the right one, I feel that there is still a lot that can be learnt if we followed the more classical tradition where economics was a social theory yet it was not a form of a “replicative process.” The dire state of DSGE should not be a reason to separate theory from policy as it should be a good reason to expect better economic theory.