La recensione di Why Economics Works, When It Fails, and How To Tell The Difference di Dani Rodrick a cura di Unlearning Economics (4 settembre 2016) discute dei limiti dell’economia neoclassica della quale l’autore espone i diversi pregi che invece avrebbe.
Dani Rodrik’s book Economics Rules: Why Economics Works, When It Fails, and How To Tell The Difference is probably the most readable and persuasive book among the economists-defending-their-discipline genre. Rodrik aims to rebuff common criticisms of economics: that it is overly abstract and irrelevant; insufficiently interdisciplinary or pluralist; and inherently biased towards ‘free market’ policy conclusions. At the same time, Rodrik acknowledges the force of some of these criticisms and calls on his fellow economists to address them. Economics Rules is probably the most accurate summary of modern academic economics, which is often reported inaccurately, and for that reason it is a useful addition to the debate and should be widely read. However, it suffers from the fundamental flaw of defending the indefensible idea that neoclassical economics is the only valid approach to economics.
Rodrik’s argument is simple but powerful: economics is not one model but a collection of models. Models provide rigour, logic and concision in a complex world and allow economists to pin down key causal mechanisms and make predictions about some variable, policy or system. The art of the economist is to figure out which model is best suited to a given situation, and models enable this process by making clear the assumptions on which a prediction depends. Problems arise when some economists mistake a particular model for the model, confusing the abstract world of economics with reality. This causes them to become too wedded to particular ideas about how the world works and subsequent policy prescriptions (especially those economists and models of a ‘free market’ persuasion).
Though he doesn’t claim so himself, Rodrik’s methodological approach could be considered a more sophisticated restatement of Milton Friedman’s famous paper The Methodology of Positive Economics, which similarly sought to defend economic models from charges of unrealism and irrelevance. While Friedman argued that the unrealism of a theory’s assumptions does not matter as long as the theory makes correct predictions, Rodrik adds nuance to this by stating that while unrealistic assumptions are in general necessary and useful, some assumptions are so important that they must be amended to be more in line with reality. Rodrik calls these ‘critical assumptions’, stating that “an assumption is critical if its modification in an arguably more realistic direction would produce a substantive difference in the conclusion produced by the model.” By doing so he distinguishes his argument from the seeming ‘anything goes’ implications of Friedman’s essay.
Unlike Friedman, Rodrik’s book is replete with examples of economic models making falsifiable predictions and being used to address real world problems in practice, which makes it far more concrete and convincing. But like Friedman, Rodrik fails to fully define and unpack some of the crucial terms and distinctions on which his argument depends, meaning that his argument is based on similarly shaky foundations. For example, using Rodrik’s definition of a critical assumption above I could come up with a simple objection to virtually every economic model: agents do not optimise. Generally speaking, the conclusion of a given model will not follow without optimisation, which by Rodrik’s own definition makes it a critical assumption and thus one liable to be challenged.
It may seem trite to object to economic models because ‘agents do not optimise’, but there is an important point to be made here. Models impose a particular structure on the world, one which may or may not be an accurate representation of reality. Models can only make predictions, highlight causal mechanisms and generally clarify thinking if the structure they impose in some way resembles the structure of the real world. This is not true only for assumptions such as perfect information and perfect competition, which as Rodrik points out can easily be changed, but also for the more fundamental properties of economic models such optimisation and equilibrium.
This point may seem obvious but it is largely lost in Economics Rules, which seems to equate the proposition that some causal mechanism or prediction is possible with the proposition that it can be shown using a (neoclassical) economic model. Rodrik argues that models can be compared to laboratory experiments, as both models and experiments represent artificial environments, even if the former is mental while the latter is physical. But while we can be sure that the underlying physical world remains the same inside and outside the laboratory, we cannot be sure that the hypothetical world in which our model resides is the same as the world outside our minds. Once more, Rodrik does not spend enough time exploring the implications of this distinction.
Unpacking Economic Models
When it comes down to it, most economic models are simply optimisation problems, and model-based arguments show how these optimisation problems respond to various changes in their structure and parameters. It is not enough to call these optimisation problems ‘agents’ and insist that because these so-called ‘agents’ respond to the parameter changes in certain ways, real world agents – firms, consumers, governments – will do the same. It has to be proven empirically. Yet economic models typically contain variables which are not directly observable, do not have clear scientific units and so cannot be measured quantitatively with certainty. Even those variables which are in theory measurable, like inflation and unemployment, are always fraught with measurement issues.
Without this discipline, direct tests of models become difficult and the relationship of the models to reality becomes looser and open to interpretation. And with so many models to choose from to address a given problem, the choices between models – as well as the modelling choices made within each model – are often justified by arguments from outside the models themselves. Thus Rodrik’s approach does not remove the need for judgment, verbal argument, rhetoric and intuition; it simply moves it elsewhere. Rather than debating the how capitalism works and the implications of policies, economists debate the foundations and interpretations of models of how capitalism works and the implications of policies. This is not contradictory, but it does mean that the models are often as vulnerable to a qualitative objection as any non-model argument. And we should ask what exactly the models add to the debate when it is generally not provable that the structure they impose on the world corresponds to the structure of reality.
So what is the point in economic models? In my opinion, there are two main reasons to build a (mathematical) model. The first is to communicate an insight or mechanism that is either not obvious or prohibitively complicated to explain without mathematics. The field of economics contains many quantitative, interdependent components and using math can help us to understand these better than verbal logic alone. But I am sceptical that most models do this; generally speaking, in economics the insight seems to precede the model. For example, as Edward Hadas has pointed out, Paul Romer’s ‘endogenous growth models’ sought to show how technology and ideas can produce growth simply by embedding a mechanism by which technology and ideas drive growth into the model’s assumptions. This adds virtually nothing to our understanding of technology, ideas or development.
The second purpose of models is to make – or at least help to inform – a clear (maybe quantitative) prediction of how some policy or variable will perform in the future. It goes without saying that this is much harder than the first purpose, particularly in macroeconomics, though there are numerous examples which suggest it is possible: Daniel McFadden’s random utility models predicting transport usage; Steve Keen’s Minsky framework and his prediction of the 2008 crisis, Wynne Godley’s Stock-Flow Consistent models and his prediction of the same; matching models which tell us why and when school allocation mechanisms don’t work (and how to fix them). But once more, these examples are the exception rather than the norm, especially since the ‘predictions’ of most economic models are hard to separate from their assumptions (see above).
What all this means is that, generally speaking, showing something can happen within an economic model is not the rigorous litmus test that Rodrik and other economists seem to think, because it only”¦shows that it can happen within an economic model. To speak – as Rodrik and others do – of how the modelling practices of the mainstream ‘discipline’ your thinking is merely an exercise in circular thinking, as ‘discipline’ is not defined independently of the use of these modelling practices. It may be tricky to build and solve economic models, but we shouldn’t mistake this for evidence that what is being done is worthwhile.
Rodrik argues that models help to clarify areas of disagreement, but even this is not obvious. Imposing a model onto a problem may smuggle in hidden assumptions and obscure the very issue it intends to illuminate. For example, Rodrik argues that the debate over the use of monetary and fiscal policy in a recession is “essentially about whether recovery is hampered by the economy’s demand curve or supply curve.” In other words, activist policy is only justified if you believe the economy has a demand problem, not a supply problem. But this assumes that demand and supply are independent of one another, and that supply is unrelated to activist fiscal or monetary policy, an assumption which is disputed by Kaldorian demand-led frameworks and which has recently been questioned by the Federal Reserve. And this is before we even get into the different effects fiscal and monetary policy may have on debt levels, which the demand-supply model also excludes.
On other hand, and perhaps surprisingly, I actually agree with much of what Rodrik’s praises about mainstream economics: clarity about what you are saying, what it means and how it is different to what has been done before. Heterodox economics and mainstream economics both have their fair share of good and bad, but there is little doubt in my mind that mainstream research is generally much clearer. However, it is my opinion that these things stem from good professional practice, not from the models themselves. You can present a neoclassical model with clarity so that everyone in the room knows what you meant; you can present the same model terribly. The same goes for verbal argument, or qualitative evidence, or any non-mainstream mathematical technique. If economists spent at least some of the time they currently spend on a particular branch of maths honing their skills in these other areas, they might realise how much they have to gain from them, and how much of the time the models are not necessary. This leads me to the other side of Rodrik’s defence of mainstream economics: his rejection of other approaches.
If Rodrik is at his strongest when discussing particular neoclassical models and their applications, he’s at his weakest when discussing non-neoclassical and non-economic approaches. Just as his discussion of the former benefits from a broad array of concrete examples, his discussion of the latter suffers from a failure to discuss any examples, coupled with a series of sweeping, unsubstantiated assertions about what is wrong with them. Thus, despite the fact that Rodrik considers himself “well exposed to”¦different traditions within the social sciences”, one is forced to conclude that he is almost completely ignorant of what they – as well as non-neoclassical economics – have to offer.
For example, Rodrik dismisses calls for methodological pluralism in economics with the bizarre claim that “no academic discipline is permissive of approaches that diverge too much from prevailing practices”. This is patently untrue: pluralism is simply par for the course in every other social science and almost every other discipline. As Alan Freeman once commented, economics is actually “more committed to the unity of its doctrines than theology, whose benchmark simply states that ‘Much of the excitement of the discipline lies in its contested nature.'” There is even a case that hard sciences like physics are more pluralist than economics. The insistence on a particular, rigid, deductive mathematical framework for approaching issues is a characteristic of economics and economics alone.
Most strangely, the case for economic pluralism follows directly from the defining theme of Rodrik’s book: that economics is about choosing the best model for the best situation. If this is so, how can he dismiss non-mainstream models outright? This would entail either a convincing methodological demonstration of why these models can’t be used in general, else a series of case studies of why popular non-neoclassical models are inferior to neoclassical ones. Without engaging with pluralist literature – model-based or otherwise – Rodrik has no clear reasons as to why he can reject such approaches. Personally, I would love to see some engagement from mainstream economists with ideas like Minsky’s Financial Instability Hypothesis; capacity utilisation models; agent based models; cost-plus pricing; endogenous money. Note here that ‘engagement’ does not mean building one model which superficially claims to incorporate these ideas, then carrying on as normal.
Similar weakness is evident in Rodrik’s discussion of other social sciences and his (largely unfavourable) comparison of them to economics. He makes another bizarre claim: that it would be “truly rare” for junior researchers to challenge more senior researchers in other disciplines, with his only citation as the Sokal affair, where a nonsense paper was accepted into a sociology journal. It is not clear how this is directly relevant, but presumably Rodrik’s argument is that the editors simply accepted the paper because it referenced and agreed with them. But similar controversies have happened in many areas such as physics and electrical engineering. Are these just woolly, “interpretative” disciplines too? This really says more about the modern peer-review process than it does about any particular discipline. Besides, research suggests that economics is one of the most hierarchical disciplines in the social sciences.
It is a shame that it has become fashionable to lazily dismiss social sciences, because even a cursory knowledge of these disciplines shows that they have a lot to offer. Read the Philosophy Journal Analysis and tell me verbal argument can’t be clear and concise (entries are limited to 5 pages, in contrast to the 30-40 page entries that typically dominate economics journals). Read the socio-economic review, or sociological science, and tell me sociologists have nothing interesting to say about economics and statistics. Read the psychologists who have developed a simple procedural rule for individuals evaluating lotteries which outpredicts the most complex mathematical models economists have come up with, and tell me that economists are better at making empirical predictions.
All in all, once you actually engage with non-neoclassical economics and other disciplines, it is simply not credible to claim economics has any kind of theoretical or empirical superiority.
Though I have been critical of Economics Rules in this review, I regard it as an extremely readable, interesting and valuable book. Rodrik’s summary of mainstream methodology is undoubtedly the most up-to-date in a debate which often seems stuck in the past. His sections on international trade and development are, as anyone who is familiar with his work should know, unfailingly excellent. But I can only regard the mentality that Rodrik (and most of the profession) seem to share – that the ultimate test for economic arguments consists of shoehorning them into neoclassical models – as intensely frustrating, especially when it is combined with a complete lack of engagement with other approaches. For this reason, it must be concluded that the only thing economics rules is the abstract world it has created for itself.
 More precisely, I could say that preferences are not complete, which would mean that they could not be represented by a utility function and therefore that the agent would not be able to choose their (maximising) most preferred set.
 I’ve actually reverse engineered a conclusion I wanted from a model myself, and if I can do it then anyone can.