Even many mainstream economists now concede that macroeconomics needs a major rethink. The Great Financial Crisis and the pandemic emergency have pushed actual policy far beyond the parameters of orthodoxy, while real developments have persistently challenged central tenets of older thinking, such as the supposed close connection between the money supply and inflation.
Lance Taylor and Nelson Barbosa-Filho (2021) illustrate the problem by explaining how a cost-based structuralist approach can provide an account of inflation which is more coherent than the conventional monetarist approach and more policy-useful, highlighting a resulting policy trilemma. They identify much of the incoherence of the mainstream approach with ‘ontological blindness’, as evidenced by the persistence of the rational expectations assumption. In other words Taylor and Barbosa-Filho argue that rational expectations do not make sense in relation to the real world.
Mainstream macroeconomics has relied heavily on Dynamic Stochastic General Equilibrium (DSGE) models, which incorporate rational expectations. But the failures of DSGE models were laid bare in the 2008 financial crisis. They had failed to predict the crisis. Further, by not even allowing for the possibility of crisis they were ill-equipped to explain what had happened and therefore also to generate policy solutions. Here we pursue further the charge of ontological blindness as it applies to DSGE models, and the idea that considering the future of macroeconomics requires us to start at the ontological level.
In the wake of the crisis, non-mainstream economists continued to articulate their longstanding critique, at the foundational level, of general equilibrium theorising as it applied to DSGE models. At the same time a series of powerful critiques of DSGE models focused on model structure. In particular the reliance on the representative agent was shown to be unduly restrictive, particularly in terms of diverse sets of beliefs. The absence from these models of a financial sector was particularly problematic when considering a financial crisis.
The dominant reaction within mainstream macroeconomics was paradigm defence. In an attempt to rectify the shortcomings, economists tried out a wide range of adaptations: developing heterogeneous-agent models, incorporating financial structures, exploring expectations in terms of beliefs, and so on. The result was explanations for the crisis in terms of a range of market frictions and belief shocks. The policy implications that followed, e.g., that financial frictions should be removed and transparency increased, can be seen as an attempt to make the real world more like the model. But DSGE models were still not able to deal adequately with substantial departures from equilibrium other than by adaptations after the event.
Although the efforts at paradigm defence seemed to have been successful in that the force of the criticisms of DSGE modelling did not seem to dent its hegemonic position, especially among central banks, the cracks showed. Consideration of alternative approaches was seen to be warranted. David Vines and Samuel Wills are to be congratulated for taking up this challenge. They brought together a diverse group made up mostly of mainstream macroeconomists to explore and address head on the criticisms of DSGE models, and to consider future directions for macroeconomics. The outputs of this Rebuilding Macroeconomics Project are set out in the Oxford Review of Economic Policy (2018, issues 1-2 and 2020, issue 3).
Of particular interest is their discussion of different types of models other than DSGE models, notably the structural econometric models (SEMs) which play a major role in informing policy debate. Such a discussion could form the basis of moving towards macroeconomics without DSGE. But, while the balance of opinion in the 2018 papers was in favour of a plurality of models, they did not envision any paradigmatic shift from DSGE models at the core. Indeed other types of modeling were seen as enabling the development of a more comprehensively core DSGE model.
The new aspiration arising from the 2020 round of papers is to downgrade DSGE models from core models to ‘toy’ models whose role would be to allow model builders to take a ‘quick first pass at important questions’. The aim of these DSGE models ‘is to provide the simplest possible way of seeing how the fundamental pieces of the economy fit together, subject to it being “properly microfounded”’ (Vines and Wills 2020: 430). Further, in an effort to address the macroeconomic problems which had been identified as deviations from equilibrium, the emphasis is now put on developing multiple-equilibrium DSGE models to account for the different paths that an economy might take. This effort would be supported by multiple small toy models and by much greater emphasis on structural econometric models.
But why are DSGE models still in the mix at all, and in a key position? Given all the criticisms, what can such models tell us, even as a ‘first pass at important questions’? Multiple equilibria do allow for discussion of a wider range of scenarios, but any discussion of a particular scenario is still constrained by the requirements of general equilibrium theory. These requirements are at the root of the more fundamental critiques of DSGE. While Vines and Wills set out an impressive research agenda to flesh out this multiple-equilibrium approach, we need to reflect on the constraints imposed by general equilibrium theorising itself.
We therefore need to revisit the fundamental problems with general equilibrium theory and the restrictions it imposes on what is admissible. Individual behaviour needs to be determinate such that indeterminacy can only enter due to an ad hoc restriction or else as a shock. Institutional structures need to be fixed, or else evolve in a deterministic way. Further, the very focus on equilibrium as an outcome of market forces severely constrains the subject matter. This is epitomised by the treatment of uncertainty as risk, or concealed risk. Without being able to absorb the significance of fundamental uncertainty for individual behaviour, for social structures and institutions, and for economics itself, the subject matter and the theoretical structures designed to capture it are inevitably constrained.
The force of these restrictions is embodied in the requirement for microfoundations, whereby all propositions need to be derivable from axioms with respect to individual behaviour. The issue of microfoundations (their role and content) lies at the heart, not only of critiques of general equilibrium theory but also of debate within mainstream macroeconomics itself.
One position is to double down on the microfoundations agenda. The New Monetarist approach is an exemplar of applying the microfoundations logic most consistently to DSGE modelling. Thus all those elements deemed to be missing from DSGE models, such as financial market frictions and diverse, shifting beliefs, need to be generated endogenously on the basis of the rationality axioms. It is hard not to see this endogenisation within a core DSGE model as the single-equilibrium end-result of the proposed multiple-equilibrium agenda applied to DSGE models.
Yet it is a real question whether such a goal is achievable. It requires social structures and social aspects of behaviour (such as trust), and indeed of identity, to be explained by rational self-interest. If altruism is to be included, that too needs to be incorporated in a deterministic fashion. It requires institutional structures likewise to have a reductionist explanation. Even more challenging is the prospect of explaining the evolution of behaviour and institutions (deterministically) in terms of rational self-interest. If structures and behaviour evade deterministic representation then a different methodological framework is required.
Given the difficulty in building a fully microfounded theory which can address data, the need to address urgent policy needs has led many modellers to proceed largely without microfoundations. Simultaneous Equation Models (SEMs) in particular have become increasingly popular. It has been seen as a regrettable necessity that SEMs are ‘inconsistent’ with the microfoundations of DSGE models. But SEMs have the great benefit of ensuring analysis that accords with evidence and is thus of direct usefulness to policy-makers. The Rebuilding Macroeconomics Project exploration of different types of mainstream macroeconomic models supports parallel, and indeed integrated, development of DSGE models and SEMs.
But what is the logical case for integrating inconsistent models? What inconsistency means depends on the epistemological context. Within a closed classical-logic deductive system like general equilibrium, inconsistency implies incoherence. So how can SEMs operate alongside, and indeed be derived from, DSGE models, when they are inconsistent? Vines and Wills (2020) go some way to addressing this question and recognise the scale of the challenges it raises. But there needs to be some epistemological grounding for this discussion. And that epistemological grounding needs to be articulated with respect to a declared ontology.
Inconsistency means something very different in the open-system epistemology that is common outside of mainstream macroeconomics. This epistemology is designed specifically to address an open-system ontology, which cannot be captured in an all-encompassing formal model. It privileges ontological consistency over internal consistency. Knowledge is then built up from multiple strands of reasoning, some – but only some – of which may involve formal mathematical models. Taylor and Barbosa-Filho (2021) is an exemplar of such an approach.
From a standpoint like this, the first important implication is that any formal model is only one of several contributors to theoretical argument – no one model can be expected, as in mainstream theorising, to provide core answers to policy questions. This applies equally to DSGE models and SEMs. Second, different strands of argument may well take different starting points and in that sense are thus inconsistent. For example, Keynes analysed the forces generating the money supply in one analysis and took the money supply as given in another. Indeed in macroeconomics it may be that no part of the analysis starts at the micro-level. King (2012) makes a powerful case against the need for any microfoundations (or for that matter macrofoundations) at all.
Since the attraction of SEMs is their greater closeness to evidence, they highlight the particular issue of realism of assumptions. In the debates that followed Friedman’s argument against the need for realistic assumptions, it was pointed out that assumptions can be unrealistic in a variety of ways, some potentially more problematic than others. In particular, there is a distinction between a simplifying assumption and a fictitious assumption. Keynes’s assumption of a given money supply is an example of the first. It is straightforward to relax it in further analysis (having kept a theory of money supply at the back of one’s head). But the rationality axioms in their traditional form refer to fully-informed, rational, atomistic, self-interested agents. This involves assumptions which are not simplifications, but rather fictions. If they are relaxed, the whole structure of DSGEs is affected. There is no separate analysis behind the assumptions other than arguments for changing them altogether. So what are we to make of DSGE models based on such assumptions in relation either to different small toy DSGE models or to SEMs which don’t make such assumptions?
On the face of it, it may seem that the Vines-Wills proposal for a plurality of types of models is very like the open-system epistemology outlined above. The idea of smaller models designed to focus on particular aspects of the macroeconomy, which will guide its future path, seems like a positive development. But there are crucial differences. First, the goal is to arrive at a DSGE synthesis informed by evidence from policy models. The formal DSGE model remains central to analysis rather than one of many contributions to a broader analysis. Instead, in an open-system epistemology, indicating the likely effects of policy requires analysis of forces behind institutions and behaviour beyond the scope of statistical forecasting. Second is the absence of a common ontology to guide the content of, and interaction between, models. This is important for devising a strategy for incorporating inputs from policy models, but also for guiding assessment of the relevance at any one time of one or other of the multiple equilibria. But thirdly there is the stumbling block which is general equilibrium methodology itself. As we continue to see, this deductivist methodology seeks to absorb all other types of analysis into a complete synthetic structure.
Exploring different approaches to theorising and of relating theory to evidence is a constructive way to address the past failings in macroeconomics. This note is offered as a contribution to that exploration. It is to be hoped that the Rebuilding Macroeconomics Project continues its efforts to explore the potential for paradigm change. This would involve exploring the rich ground of theorising beyond DSGE models, to the extent of considering a future for macroeconomics without DSGE models – with or without multiple equilibria. The starting point needs to be explicit statements about the nature of the real world.
King, J E (2012) The Microfoundations Delusion: Metaphor and Dogma in the History of Macroeconomics. Cheltenham: Edward Elgar.
Taylor, L and Barbosa-Filho, N (2021) ‘Inflation? It’s Import Prices and the Labor Share!’, INET Working Paper No. 145, 20th January.
Vines, D and S Wills (2020) ‘The rebuilding macroeconomic theory project part II: multiple equilibria, toy models, and policy models in a new macroeconomic paradigm’, Oxford Review of Economic Policy, 36 (3): 427-97.