New book on dynamical, agent-based models for economics

[UsableThought]

This may be of special interest to mathematicians, physicists, and anyone who is skeptical of the various "social sciences."

The book is The End of Theory: Financial Crises, the Failure of Economics, and the Sweep of Human Interaction, by Richard Bookstaber; 2017, Princeton U. Press. I've only just started reading this; so far I'm enjoying it immensely, for Bookstaber is unexpectedly funny, irreverent, and sharp in covering a topic that often seems humorless, conforming, and dull: namely, economics.

His thesis is that (1) all the usual theories of economics - including the neoclassical school whose precepts often get approvingly cited in discussions in this forum as if they represented reality rather than merely being part of a model - are incapable of predicting the emergence of crises (apparently many economists agree with this part); (2) that where these models go wrong is in simplifying economic life by assuming it is entirely made up of rational actors who have full knowledge of markets; and finally (3) the way toward better modeling is to adopt an agent-driven, dynamical approach that has some similarity to certain problems in physics with computationally irreducible systems/problems, with the agents designed so that they are not rational and do not have full knowledge of the collective system but rather, follow context-dependent heuristics, thus more closely resembling humans and human-driven institutions. More on dynamical systems here: https://en.m.wikipedia.org/wiki/Dynamical_system

A way to summarize his argument is that accurately predicting economic events - especially crises, where models are put to the most severe test - is similar to predicting a complex event such as the actual paths traversed by the Voyager probes. From a 2013 EDN article about the "mathematics of space travel":

The problem is predicting exactly how the gravity of the Sun and a planet will influence Voyager’s trajectory on its journey out of our Solar system. Renowned astronomers have struggled with this problem for at least 300 years . . . Minovitch had access to the IBM 7090 computer, unlike Newton and others, and endeavored to use the method of mathematical iteration to craft a solution that was not realizable without the processing power of such a computer.​

Minovitch understood that the only way to completely describe a system of 3 bodies is to observe how their behavior unfolds. This is where the number-crunching power of the IBM 7090 came into play.​

For economics, the difficult difference would seem to be defining the heuristics for humans as economic agents; not quite the same as modeling celestial bodies, at least at first impression. Yet surely assumptions have to be made and tested for models of Voyager etc.; so there should be at least some developmental overlap.

I haven't yet looked very far for reviews; I'm sure many will be highly critical as Bookstaber is likely attacking the vast majority of professional economists. But he is no fool, and I already have the sense, from the little Googling I've done, that some in the profession realize he has a point & are quite interested. The computational power to carry out his proposal (which I am guessing he is not alone in making) is certainly there.

I searched PF and Bookstaber's name has only come up once previously, in a reply @ZapperZ made to a thread inquiring rather vaguely whether any developments in economics have referenced physics: https://www.physicsforums.com/threads/applying-physics-to-economics-finance.897391/#post-5644898

Even in the first few chapters, Bookstaber does indeed refer to quandaries in both mathematics and physics; e.g. he reminds the reader that no less a personage than William Stanley Jevons, one of the earliest and foremost champions of the neoclassical model, himself lamented that the three-body problem had no general solution in physics - thus making it unavailable to the new kind of formula-driven economics he was inventing as what he believed would be a final solution to - yes - the prediction of economic crises (in his time, having to do with railroads). And Bookstaber summons other math/physics concepts as well - not in passing or as metaphors, but as concrete guides to developing agent-based models.

Here is a link to the publisher's page for the book; the first chapter is available as a PDF and I recommend it to anyone interested: http://press.princeton.edu/titles/10972.html

NOTE: I see that PF's excellent automated facility for offering up similar links has found one titled "Agent-Based Simulation for Political Sciences" - alas it got only two replies and relates to political systems exclusively. Even so this suggests that Bookstaber (what a great name, by the way) isn't the only one interested in modeling of this sort for human-driven systems. My hunch is that deeper in the book he will point to research in this area.

 

[256bits]

This may be of special interes to mathematicians, physicists, and anyone who is skeptical of economics as a "scientific" profession

Chaos theory, Mandelbrot, wave theory, ..., have all been attempted to apply to economics over and above the basic "many sellers-many buyers" introduction to the field. Supply and demand is what most people could espouse as being their understanding of "economics," and then suffer glazed eyes when delving further into more a more in depth analysis, as you have noted. If .Bookstaber, from your short review, is a good read, I may look into it. Nothing better than a subject that can be made interesting by the author.

here is just one adaptation of a science/mathermatics adaptation to economics.
Mathematically, Game theory economics has a long succession of followers.
https://en.wikipedia.org/wiki/Game_theory
Noble prize in 2005
http://www.nobelprize.org/nobel_prizes/economic-sciences/laureates/2005/press.html

Game theory is "the study of mathematical models of conflict and cooperation between intelligent rational decision-makers". Game theory is mainly used in economics, political science, and psychology, as well as logic, computer science and biology.[1] Originally, it addressed zero-sum games, in which one person's gains result in losses for the other participants. Today, game theory applies to a wide range of behavioral relations, and is now an umbrella term for the science of logical decision making in humans, animals, and computers.

https://en.wikipedia.org/wiki/John_von_Neumann

 

[Borek]

I feel like huge part of the problem is that while we don't know general solution to 3 body problem, we have equations that are known to describe it precisely (so not being able to solve them is just a technical problem). I doubt such equations exist for economics and society - or, rather, even if we can write such equations, they can describe only some kind of 'averaged' reality, while the 'real' reality is much more chaotic and can react to random, unpredictable events (say, stupid typo in a message goes viral and ends with a panic on the stock exchange).

Perhaps it is just a difference between continuous and non-continuous systems?

Nonetheless fascinating subject.

 

[UsableThought]

I doubt such equations exist for economics and society - or, rather, even if we can write such equations, they can describe only some kind of 'averaged' reality, while the 'real' reality is much more chaotic and can react to random, unpredictable events (say, stupid typo in a message goes viral and ends with a panic on the stock exchange).

Outlandish freak events seem to be part of Bookstaber's attack on conventional economics. He uses the phrase "radical uncertainty" for those things we not only don't know, but don't know we don't know (to borrow from Donald Rumsfeld) and thus can't include in a model. Beyond that I haven't read far enough yet to see what he proposes, e.g. for the sort of example you give.

 

[fresh_42]

In my opinion this is a long overdue point of view. However, the actual price building depends on so many variables, that I have some doubts a single model could cover them. There are: balance press conferences, analysts' visits, phone calls with fond managers, private "lunch talks" between traders, automated stop loss orders, synthetic products like options, secondary markets, various (and different) risk systems of deciders and stock holders, computer trading itself, change of indices, normal press conferences, political decisions, inter-dependencies, natural catastrophic events (just read that the Vanilla price is expected to increase due to a storm in spring on Madagascar) and probably many more. In real life markets have more in common with Casinos than with rational decisions. I think generalizations and bundling of variables cannot be avoided, which automatically bears the risk of uncertainties. Maybe it's far more like physics than the author might think of: different models for different ranges of variables. I've always found the boundary conditions made in economical textbooks were far too idealized. On the other hand, I never understood chartists. It shows what people do to get a hold on complex systems. Successful investment is often an equivalent to someone's perfect gut feel, which in turn means that all efforts to calculate optimal solutions, and there have been many, are doomed to fail.

 

[StoneTemplePython]

People at the Sante Fe Institute have been working on this sort of thing for quite some time now. Last I checked these dynamical agent based models don't offer particularly useful predictions.

Typically when you point out how bad predictions are in macro econ you get strange responses -- people either point you to some other methodology (that also has bad predictions) or retreat behind this odd line that econ is supposed to be normative not positive (which is even stranger when you start peering at the underpinnings of normative economics)

 

[Asymptotic]

I'll get a copy of 'The End of Theory' next time I go on a book-buying jag. Thanks!

Two other authors you may find of interest: Mark Buchanan often refers to the inadequacies of standard economic theory in his books and other writings, and economic factors comprise one leg of the stool Peter Turchin has fashioned for his cliodynamics models.

 

[256bits]

People at the Sante Fe Institute have been working on this sort of thing for quite some time now. Last I checked these dynamical agent based models don't offer particularly useful predictions.

That's kind of what I thought too.
They don't seem at to be the next best thing to baked bread, economically speaking.

 

[UsableThought]

People at the Sante Fe Institute have been working on this sort of thing for quite some time now. Last I checked these dynamical agent based models don't offer particularly useful predictions.

That's kind of what I thought too.
They don't seem at to be the next best thing to baked bread, economically speaking.

I speak from ignorance, but it seems important here to distinguish between the approach as a concept, versus its slow evolution (assuming it is evolving) toward practical application. The Santa Fe Institute itself has only been around since 1984; three-plus decades may seem a long time but for difficult new directions can be summed up as "early days."

Me, I'm as much interested in the dissection of conventional economics as I am in the alleged advantages of agent-based systems. Of course seductive writing can hide a lot of intellectual sins.

If you can get past the pay wall, a mostly positive review in The Economist concludes with what the reviewer believes to be the book's strengths and limitations:

The analysis is top-notch, and anyone who wants to understand the workings of the financial system will benefit from reading this book. But those looking for a quick fix will be disappointed. Mr Bookstaber says there is no specific model to deal with crises. Instead, he is describing a process – an intellectual approach to understanding the system.​

Furthermore, although the author gives a kicking to mainstream economics in general, his analysis focuses entirely on the financial sector. The problems that bedevil economists (inflation, unemployment, productivity) do not feature. The challenge facing traditional economists is to incorporate Mr Bookstaber’s insights into their forecasts. A daunting task.​

 

[f95toli]

I feel like huge part of the problem is that while we don't know general solution to 3 body problem, we have equations that are known to describe it precisely (so not being able to solve them is just a technical problem). I doubt such equations exist for economics and society - or, rather, even if we can write such equations, they can describe only some kind of 'averaged' reality, while the 'real' reality is much more chaotic and can react to random, unpredictable events (say, stupid typo in a message goes viral and ends with a panic on the stock exchange).

An even more serious problem is that is that many people (who should know better) don't realize that there is a problem. That is, they think the models describe reality and that any discrepancies just means that you have to tweak the model a little bit. This is one of the "long tail" problems in economics where extreme events are discounted because they are improbable.

Don't get me wrong, models are obviously extremely useful and we do need them; but it is also important to understand the limitations of the models (in economics as well as in every other field); and this is not always the case.
What seems to have happened in economics is instead that the simplifying assumptions (e.g. fully rational agents) was -at least in part- turned into political ideology.

 

[UsableThought]

What seems to have happened in economics is instead that the simplifying assumptions (e.g. fully rational agents) was -at least in part - turned into political ideology.

This is an important point - and not just for the assumption of rational agents, but for many other assumptions carried by one model versus another.

Here's a pithy summary of the problem, taken from the heading of the Wikipedia page that lists sub-pages for "Economic ideologies":

Economic ideologies are here defined as moral positions - how economies should be structured, as compared to economic theories - systems of propositions concerning how economies work. Of course, the two are related. Certain theories support or preclude certain ideologies. For example, someone who believes in the Keynesian theory of economics is quite unlikely to subscribe to libertarian ideology.​

The key to ideology is this taking of "moral positions." When a model's assumptions are treated as bedrock realities, usually this will be followed by a great deal of moralizing. We see this in political posturing at the federal and state level; however it has also popped up in virtually every recent "General Discussion" thread that I have seen here at PF about social issues which involve economic policies or consequences.

Typically, those of us caught up in one view or another (certainly this would include me) are making assumptions about the nature of markets; these assumptions very often connect to moral beliefs about the proper responsibility of individuals versus that of society and/or government. We seem unable to separate out the inherently imperfect nature of market models from our acquired definitions of moral behavior. We each have a history in which we were taught or otherwise imbibed our moral beliefs; yet almost no one is willing to acknowledge having been subject to influences outside of our power - e.g. where we were born, where we grew up, who were our important role models, etc.

 

[256bits]

I speak from ignorance, but it seems important here to distinguish between the approach as a concept, versus its slow evolution (assuming it is evolving) toward practical application. The Santa Fe Institute itself has only been around since 1984; three-plus decades may seem a long time but for difficult new directions can be summed up as "early days."

Me, I'm as much interested in the dissection of conventional economics as I am in the alleged advantages of agent-based systems. Of course seductive writing can hide a lot of intellectual sins.

If you can get past the pay wall, a mostly positive review in The Economist concludes with what the reviewer believes to be the book's strengths and limitations:

The analysis is top-notch, and anyone who wants to understand the workings of the financial system will benefit from reading this book. But those looking for a quick fix will be disappointed. Mr Bookstaber says there is no specific model to deal with crises. Instead, he is describing a process – an intellectual approach to understanding the system.​

Furthermore, although the author gives a kicking to mainstream economics in general, his analysis focuses entirely on the financial sector. The problems that bedevil economists (inflation, unemployment, productivity) do not feature. The challenge facing traditional economists is to incorporate Mr Bookstaber’s insights into their forecasts. A daunting task.​

OK. Good post.
One challenge for the economic people is to make sense of the Philips curve, perhaps a bit simplistic at first glance, but the relationship between inflation-employment is something they spend time on. Some head scractchng on their part to make sense of it, or outright rejection of the proposition as being illusionary.
https://en.wikipedia.org/wiki/Phillips_curve
That's where I thought, or assumed, Bookstaber was leading to, how monetary policy for example, affects that aspect of the general economy and its players and their decisions.
Anyways, I had better stop now, or someone will get the wrong impression that I know what I am talking about.