Although buzzwords and phrases such as Machine Learning and Data Science are littered throughout various media concerning how they are impacting various industries, the Economics profession seems to be particularly slow and/or resistant to adopting computational methods.
The Economist went so far as to run an article with the self-explanatory title, “Economists are prone to fads, and the latest is machine learning.” Nevertheless, Andy Haldane made important references to Artificial Intelligence in his speech to the Trades Union Congress in London on the 12th November 2015.
Besides the empirical, econometric analyses which will eventually have its practical foundations shaken by the disruptive storm of Computer Science and Computation like virtually every other aspect of our lives (directly or indirectly), there is far more that Computer Science can offer to the profession but which Economists most generally are resistant to.
After all, given the fact that peoples around the world are growing increasingly sceptical of economic forecasts and econometric models, empirical economic analyses may become increasingly less useful for public policy formulation and even less trusted.
Furthermore, as we enter an era of increasingly unconventional policies (such as Monetary Policy), empirical analyses based on both contemporary and historical data will be of increasingly limited value in a world where there are fewer precedents and, even where there are some comparisons to be made, the circumstances remain significantly different.
Agent-based Modelling in the Social Sciences
J. Doyne Farmer and Duncan Foley wrote an opinion piece for Nature back in 2009 entitled “The economy needs agent-based modelling.” However, the logic can be applied not only to economic modelling but social-scientific phenomena more broadly.
Agent-based modelling provides an alternative and arguably more flexible scientific modelling paradigm to the widely-used, contemporary conception of dynamic stochastic general equilibrium (DSGE) models whilst also potentially incorporating many of the latter’s benefits.
Robert Solow, a Nobel prize-winning Professor Emeritus at MIT, blasted DSGE models in 2010 to the United States Congress, saying:
“I do not think that the currently popular DSGE models pass the smell test. They take it for granted that the whole economy can be thought about as if it were a single, consistent person or dynasty carrying out a rationally designed, long-term plan, occasionally disturbed by unexpected shocks, but adapting to them in a rational, consistent way… The protagonists of this idea make a claim to respectability by asserting that it is founded on what we know about microeconomic behaviour, but I think that this claim is generally phoney. The advocates no doubt believe what they say, but they seem to have stopped sniffing or to have lost their sense of smell altogether.”
In 2010, The Economist published an article entitled “Agents of change” which explored the potential benefits of agent-based modelling as an alternative paradigm for economists to work with. Such thoughts are no longer as far outside the mainstream as they once were though they will still take time to seep in and truly challenge and positively disrupt the status quo; an example is Arthur Turrell of the Bank of England’s Advanced Analytics Division recently publishing the article “Agent-based models: understanding the economy from the bottom-up” as part of the Quarterly Bulletin 2016 Q4.
Whatever one’s modelling preferences are, Agent-based Computational Economics is becoming more prominent and, as an alternative modelling paradigm, it is likely here to stay whether it captures mainstream economics or not.
Economic Theory Will Resurge
Many students and practitioners dismiss critics economic theory as being ‘out of touch’ with reality and inaccessible (due to the often dense and mathematical nature of the literature). Although the Journal of Economic Theory is one of the “nine core journals of economics”, it is less widely read than other ‘core journals’ and this is partly due to the density of the material but also mostly because individuals often dismiss pure theory in favour of empirical and theoretical synthesis; unfortunately, this means that empirical analyses are routinely less informed than they should be.
Nevertheless, journals such as the ACM Transactions on Economics and Computation (TEAC), Computational Economics and the Journal of Economic Dynamics and Control will become increasingly important in the quest to extend the frontiers of Computational Economics generally and Economic Theory in particular.
For instance, as we live through times of unprecedented economic policy (the nature of contemporary monetary policy being a prime example), empirical analyses and historical case studies provide information of increasingly limited use in an ever-evolving, dynamic and complex economy.
Therefore, a recourse to theory and the imagination is necessary – the best way to visualise, construct and understand the complexities of various economic theories is to indulge in the application of computational methods and simulations to make it accessible to people who may not otherwise bother to simply read the mathematics and follow its implications (your author being routinely guilty of this).
Furthermore, the application of insights from economic theory and computational methods go even further beyond what we might conventionally perceive as ‘core’ economic policy.
An example of this is the work of Francisco C. Santos at the Department of Computer Science and Engineering at the University of Lisbon in Portugal; a major theme of his research includes the incorporation of computational (evolutionary) game theory and population dynamics to simulate climate change governance outcomes.
Investing in Computational Economics
The resources devoted to the development and application of computer science to particular fields will always overshadow others in terms of the potential for their usage. For instance, life sciences applications in University Departments are of particular interest due to their potential to generate novel treatments, improve them or generate/increase research-based income.
Both in the UK and the USA, investment and interest in Computational Finance has generally outpaced investment and interest in Computational Economics. In the UK, for example, there is a Centre for Doctoral Training in Financial Computing & Analytics which is a collaboration of heavyweight academic partners such as University College London, the London School of Economics and Imperial College London which is also supported by heavyweight financial services institutions.
Similarly, Computational Finance masters programmes are available across the UK at top institutions such as the University of Essex, the University of York, the University of Exeter and many more. Computational Economics, however, has been slower to catch on in British Computer Science Departments.
This is partly because Finance generally has more money available and the demand for these skills and research feedback into research agenda priorities due to funding availability from industry, for instance.
In the UK, the largest University-based research group in Economics and Computation is at the University of Liverpool’s Department of Computer Science (where the author is a Masters student) which sits within its large Algorithms research section (the other major research section being Artificial Intelligence).
Projects in the Economics and Computation research group include ‘a Game Theoretic Analysis of the Space Debris Removal Dilemma’ as well as researching ‘Efficient Algorithms for Mechanism Design Without Monetary Transfer’, and ‘Algorithmic predictive analysis in customers utilise spending and supply chains’.
They specialise particularly in Algorithmic Game Theory and Mechanism Design and, indeed, recent research from Professor Tuomas Sandholm at Carnegie Mellon University and grad student Noam Brown where they designed an AI which used game theory to beat four top players at no-limit Texas Hold ‘Em is further testimony regarding the potential for Algorithmic Game Theory and Computational Economics in terms of yielding useful applications.
Although the algorithmic game theory and mechanism design research communities are smaller compared to Liverpool’s, other prominent researchers in the UK are based at the University of Warwick’s Division of Theory and Foundations, the London School of Economics’ Department of Mathematics, and the University of Oxford. The University of Essex also has a significant Centre for Computational Finance and Economic Agents but across the Atlantic in particular that a lot of traditionally heavyweight Departments are engaged in such research.
Whether it be at Stanford’s Social Algorithms Lab (SOAL) at its Institute for Computational & Mathematical Engineering, MIT’s Computation and Economics Research Group, Harvard’s EconCS Group, UC Berkeley’s Theory Group, Caltech’s Department of Computing + Mathematical Sciences, Cornell’s Computational Social Science scene more broadly , George Mason University’s Center for Social Complexity, or the many others in the USA, leading American Universities’ Computer Science Departments all have particularly thriving and active research groups in Computational Economics and Computational Social Science more broadly.
Computer Scientists learn a lot from Economists but Economists and Social Scientists more broadly still have a lot more to learn from Computer Scientists.
Despite the significant progress made by university computer science departments in tackling these fundamental problems, university economics departments worldwide are generally lagging behind in their investment into computational economics, and the potential for the myriad of computational methods to enhance economic analyses remains under-exploited and undervalued.
In any case, as computational economists (and computational social scientists more broadly) advance the frontiers of knowledge as it pertains both to policy and industry, economics departments will have no choice but to evolve, innovate and adapt or be swept away into irrelevance due to what Schumpeter might call ‘Creative Destruction’. It is not difficult to see what rational individuals would do even if, in reality, there is a delay experienced in witnessing rational social/collective outcomes.
On the whole, computer science is poised and well-placed to play an important role in revolutionising the economics profession; it will be a gradual but necessary process because academics (like many of us) tend to be resistant to change when employing research methods and methodologies.