The End of Theory: Financial Crises, the Failure of Economics, and the Sweep of Human Interaction by Richard Bookstaber, Princeton University Press, 2017, 240 pages; $24.95
The End of Theory is Richard Bookstaber’s second book. Richard Bookstaber is a well-known and highly respected finance industry professional. His first book – A demon of our design has become a classic. The first book, published in early 2007, presciently predicted risk embedded in the global financial system because of tight linkages – something which all us saw in 2008. Richard Bookstaber has worked at Morgan Stanley, Salomon Brothers, Moore Capital, and Bridgewater as well as the Financial Stability Oversight Council and the Office of Financial Research. He is now Chief Risk Officer at the University of California.
As I see it, the second book is a continuation of his previous book. At heart of the previous book was the notion that complex systems are highly prone to non-linear response to exogenous shocks. Essentially markets are fragile because the way various inter-linkages work can change quickly and in unanticipated fashion when faced with an exogenous shock. In financial markets, this interlinkage is liquidity which quickly unraveled during the global financial crisis.
But the first book stopped well short of proposing a solution. The second book begins where the first book ends. The solution is still not there, but the author has attempted to put in place a framework to design and study the financial system in a more meaningful manner, proposing an alternative to the dominant paradigm. In some sense, the author has ruled out a solution by reaching the conclusion that we need to be on a continuous lookout for emerging flash points and keep tackling them as we move along. Don’t be surprised if you find some elements of Minsky’s financial instability hypothesis.
The End of Theory raises some important questions, provides several insights and a new approach to addressing economic questions as well – Agent-Based Modelling. In spite of that, in the end, I was left a bit underwhelmed – I expected more discussion on his agent-based modeling and if any interesting results that have come out of it so far. But I can understand his predicament – writing more on his recommended approach of Agent-Based Modelling could have the made the book more academic.
The book begins by addressing the extreme focus of modern-day economists to prove somehow that economics is a mechanical science in the same league as Physics, Chemistry, etc. It sees economists somehow harboring a stigma of economics not being given the same status as these hard sciences. I think of it as the fact that economics finally is a social science, which the author’s arguments convince me is true and should not be something to be embarrassed by. In fact, it is something we should be embracing.
The author then discusses in detail on why it is difficult for economics/finance to be treated in a similar manner as pure science. This can be identified to four broad phenomena endemic to the subject. These are – markets are emergent systems prone to building up unintended negative and positive feedback loops. An example of emergent systems is where everyone tries to sell off at the same time. This may be an optimal strategy at an individual level, but at aggregate level leads to extreme price movements and hits liquidity. Markets are non-ergodic where probabilities are impossible to define. This is a direct consequence of being human. We learn we change, and we interact on the basis of ever-changing context. These two forces – emergent systems and non-ergodic distributions – combine to create the third phenomenon of radical uncertainty, which is different from risk (Unknown Unknown vs. Known Unknown). Crisis arises because of an unknown unknown, never because of known unknown. And finally, Computational Irreducibility – the fact that our economic behavior is so complex that there is no mathematical shortcut or closed form solution. In a sense, it is an outcome of the first three. We cannot assume one idealistic, fully rational, a multi-period optimizing representative agent who represents the whole world. We need to approach the problems by simulating them in as much detail as possible.
After discussing the above rationale in detail, the author then discusses his preferred approach, the Agent-Based Modelling. The idea is to move away from the dominant ‘representative agent’ paradigm in economies to modeling different market participants and their interactions explicitly in as much detail as possible and then understand these interactions through simulations. This is the part of the book which is a bit underwhelming. After discussing the advantages of agent-based approach in detail and at depth, there is a lack of any interesting results which were found or have been used by regulators. While I can understand author’s desire not to give the book an academic look, this section could have been richer.
The key advantages of Agent-Based Modelling are that it allows for very rich modeling. Different actors in the financial world – asset managers, broker/dealers, asset owners, hedge funds, custodians can be represented separately, given their own connections to one another and their own rules. This allows for addressing some of the endemic challenges discussed. In fact, the author presents some designs of the financial systems and some examples of a simplified financial system.
In the final section, the author discusses the difference between deductive and reflexive decision making. The current paradigm assumes the decisions to be based on deductive logic, and therefore mathematical modeling being the best way for building theory. However, we know that the complexity prevailing in the world demands a reflexive decision making – where one starts with a rough map and keeps oneself flexible as more and more details become visible. Agent-based Modelling becomes a possible candidate for theory building in a world characterized by reflexive decision making.
In some sense, it can be said that the book proposes a particular kind of approach and belief system – Agent-Based Modelling. The downsides of Agent-Based Modelling are not discussed at all. One may come out of the book thinking it is a panacea for all ills. Agent-Based Modelling comes at its own costs – the number of assumptions will increase exponentially with the number of agents and how realistically one wants to make it. The detailing of the model itself comes from experience or from the past, whereas the future may be totally different.
Nevertheless, End of Theory is an important and interesting book for anyone who wants to understand what is wrong with the current modeling paradigm in economics and possible solutions to it. In between, you get pearls of wisdom as well as good references (both fictional and non-fictional) which can be pursued. Bookstaber quotes Jorge Louis Borges extensively, which has convinced me to pick up his short stories and understand his work in greater detail. Another one is William Stanley Jevons who initiated the challenging marriage of economics with mathematics.
Time to learn and get initiated into Agent-Based Modelling!