Reality Bites: Is the Era of Assumptions Coming to an End for Economists?

DSGE, IS/LM, OLG. If you are a follower of the dismal science, or have come across any news regarding economic modelling, these terms will be somewhat familiar to you. To the rest, they’re just acronyms.

Economics modelling is a funny thing. Instead of actually having specific laws, an all-encompassing model, or a grand unified theory, it has models which need to describe their specific purpose and characteristics through complicated names (and even weirder acronyms). Although not a natural science, economics could gain much by learning how scientists model the natural world. Indeed, we are finally seeing evidence of economists moving away from modelling with assumptions, to modelling what they observe. Whether this proves to be a prohibitively expensive gimmick that yields near-zero marginal returns, or a harbinger for a new line of thinking, is yet to be seen.

If the acronyms in the first sentence threw you off, they are ‘Dynamic Stochastic General Equilibrium’ (good introduction [pdf] & link to textbook example) , ‘Investment – Saving / Liquid-preference – Money Supply’ (simple & slightly more complex overviews), and ‘Over-Lapping Generations’ (Wikipedia entry & university-level slides [ppt]). The first is the workhorse model in modern economics, the second a classic stylised tool that can be used to grasp major concepts, and the last one another modelling tool that focuses on inter-temporal (between time-periods) dynamics.

But what the use of acronyms really highlights is the problem facing modern economic modelling – namely how to manage the trade-off between realism and solvability. Essentially, you could try to create an exceedingly complex, but realistic model; however, it would become ever more difficult to solve with each added layer. On the other hand, you could use the many simplifying tricks and tools (and even omissions) at hand to get a model that performs reasonably well. As long as there are no unforeseeable, cataclysmic market crashes.

I recall an evening some years ago when I started arguing with a few friends over whether or not economics was a real science; needless to say, the dinner wasn’t a great success story. But, it did raise the very point that needs to be discussed. Economics doesn’t really know what it is or what it wants to be. If you look at the economics course listing of my Alma Mater, the University of British Columbia, you will find a wide variety of courses. Some particularly mathematical (dynamic optimisation and econometrics come to mind), whilst others would be consider a bit softer (history and philosophical thought of economics). Economics is the awkward teenager, unsure of its own identity, usually trapped in the wrappings of Arts or Humanities, whilst being more mathematical than many of the sciences (perhaps a fate shared by psychology). All of this, in the end, is also reflected in the modelling aspect.

When modelling something as complex as an economy, you usually have to make simplifying assumptions and model based on empirical facts or theoretical notions. Assumptions about agent behaviour, households, and banks will simplify complex human interaction and behaviour. This will enable you to draw inference from the model. However, this is perhaps how far the modelling will go, because anything beyond this could be deemed too complicated. DSGE models – the current workhorses – still mostly assume that we are all homogenous agents (if we are lucky there might be some heterogeneity), that the banking sector can be ignored almost in its entirety, and that equilibria are reached in the markets.

Orthodox economists will have you believe that this is the right approach, i.e. the individuals who stand by the current wave of DSGE models. While these models do perhaps represent the forefront of economic modelling, since they take into account a fair amount of economic variables whilst still adhering to a theoretical basis, they still exist in an awkward phase between an imaginary and a real world. Imaginary in the sense that we are imposing conditions, such as various equilibria, on the model, instead of letting the model create these conditions, and real in the sense that models are still calibrated to fit the real world relatively well. The use of these models is justified by the fact that they perform relatively well when everything is business-as-usual. However, when it comes to black swan events [pdf example of Long-Term Capital Management’s insolvency issues in 1998] and the models fail, they fail spectacularly.

Now, what if we could actually model economies according to what we observe as opposed to what we would like it to be? Luckily we can. The Economist wrote a few years back about the prospects of Agent-Based Modelling (ABM). In short, instead of trying to impose artificial or arbitrary conditions on the system (i.e. assume that markets clear in our economies for example), we could model individual heterogeneous agents and their interactions and see how this system would evolve and develop over time. Although difficult to model, at least strides are being made [pdf].

Nor is this the only promising avenue of modelling an economy more realistically. At the UNEP Switch-Asia conference on Sustainable Consumption and Production in Bangkok, Thailand, Professor Steve Keen of the University of Western Sydney presented his Minsky software to us. In comparison to traditional tools of model building, Minsky “enables the simulation of economics models defined in terms of coupled ordinary differential equations.” Professor Keen himself described it as a platform to conduct real dynamic analysis, in particular of monetary flows in a society.

Professor Keen’s approach is built on the notion that money matters (which he even blogs about), and that we need to include, amongst other things, a realistic banking sector in our current models. Some would, however, have you believe that Professor Keen is a maddened heretic, but others defend his position on how to model a modern economy (to get the full scope on Steve Keen’s differences with Paul Krugman’s thinking, read this post. Also, this posting gives a good overview of the debate).

Regardless of what your personal economic beliefs are, Professor Keen does have an interesting approach to modelling. For an example of this, especially for the economics junkies out there, a two-part video tutorial introducing Professor Keen’s modelling programme – Minsky – and its application will be following soon.