Does anyone here know of any attempts to simulate a model economy on computer? I don’t mean just a bunch of equations for the main features of an economy, but rather simulate individuals. I would guess that nowadays one could simulate an economy with a few million individuals with ease.
And if its not being done then I’d like to suggest that it should be done, (see here and here).
I am stunned by the lack of responses. Is this not an important issue?
Economists, perhaps particularly Austrians, struggle with trying to understand fractional reserve banking and its alternatives. Surely, running computer simulations is the perfect way to test how the various systems could work.
No, it isn’t. The Austrian method is linked to praxeology and methodological individualism. Economics is a social science.
Give this a read if you are curious. http://mises.org/esandtam/pes1.asp
It’s not that the issue isn’t important, it’s just that computer simulations have extremely limited applicability. A simulation would involve a lot of assumptions and simplifications of human nature in order to create ‘agents’ and would really just become a self fulfilling prophecy of a sort.
The underlying problem is in the real world uncertainty dictates a lot of what happens, and there is no way to simulate that. There is no way to really simulate reality, it’s just too complex.
I don’t know of any relevant attempts. I haven’t had anywhere near the time to fully explore a worthwhile way to go about it, but I would approach it be to create a virtual world (virtual land, natural resources, animals, etc) with artificial humans living in it (that would unfortunately be almost nothing like real humans due to lack of processing power, and lack of knowledge of human biology). Each person would have their own subjective values which will change over time, and their actions (leisure, production, trade, etc) will be based on those ever changing subjective values.
This will never show what would happen in real life, however, as these are entirely different (and extremely simplistic) environments and lifeforms compared to us and our world. It could likely be used to demonstrate unintended consequences of central management, however. Let a real person try to achieve some ends in the virtual world through creating a virtual law, and watch the individuals change their behaviour in various unexpected ways. You could then roll everything back to before the law was created, and try something else, and compare results of each law, and compare them all to non-intervention, etc.
But like I said, they would not be humans, and you would not necessarily be able to apply anything learned from such a simulation to humans. It is impossible to simulate humans to 100% detail. The more detailed you get, the closer it may be to humans, but not necessarily. You’d have your work cut out for you if you try to determine which of countless variables, and which combinations of those variables, have the largest effects on what makes us human.
I’m not sure what such a simulation could provide other than a lesson in unintended consequences.
To demonstrate something such as fractional reserve banking may be better left to thorough step-by-step explanations containing as few people/entities as possible, with the wealth and currency of each individual/entity listed at each step, possibly (or entirely) with images visually showing each person/entity and their transfers of wealth/currency between each other over time.
I’ve done something similar in a spreadsheet (but no graphics or explanation, though that may it easier for some people to understand) to demonstrate that charging interest on loans does not cause economic collapse or loan defaults: http://spreadsheets.google.com/ccc?key=0Ak0d51fpM-tocDdqU2ZOZWxtN0ZESjQ3ZWsxRHZ4RVE&hl=en
I’ve wanted to make another spreadsheet for fractional reserve banking (particularly the Federal Reserve System), but I think it would be a bit more complex, and I don’t know all the necessary details well enough to do so, nor do I have the time.
The ultimate goal would be to how fractional reserve banking works, what the consequences of it are, convince the person that it doesn’t meet their goals/values well enough (you can’t just say that it’s “bad” or “wrong”, as that’s your subjective opinion/value…you need to show it doesn’t meet their subjective opinions/values), and that you know of a way fits their subjective values better, and how it does so. You probably don’t need a simulation for that, but some good, as simple as possible explanations through various mediums (thesis, spreadsheet, pictures, animation, etc). Different people will respond differently to the way they are presented the idea.
P.S. I just now read the two blog posts, and I will again stress that the simulated humans will be different from real humans…they will be far more simple, have a different nature (they’re in a different environment than real humans), have different problems, etc. It would be like observing very simple aliens in a very simple environment, relative to humans and our environment. The changes in our values are practically unpredictable over time, but that does not mean they are random or non-deterministic. Unless you simulate how our values are really determined (as in the the molecular processes of the brain, our environment, etc), the artificial human is of a different nature, and not human, and may result in the simulation being unreliable in demonstrating how real humans would act.
It would be extremely difficut if not impossible to simulate the behavior of one individual in the economy of today much less a bunch. Individuals have extremely complex preferences and desires that are subject to constant change. One simple example: I would be willing to pay $300 for an X Box, but having played a WII last night I now would prefer a WII at $400. But the game I played is on both so I still want the WII only if it is $100 or less more than the X Box. But my girlfriend wants to watch chick flicks on a Blue Ray system. I would buy the Sony Blue Ray game player if it is up to $150 more than the Wii. All of this could of course change if a new game comes out or if the next version of one of the systems does something I have not even considered.
Doing a computer simulation is meaningless when your assumptions are wrong.
Look up Cellular Automata and Agent-Based Modeling. They are essentially the same thing under different headings and they consist of something similar to what you are talking about. You can’t really simulate “individuals” though. Rather, CA and ABM simulate individual elements which evolve over time according to a predefined set of finite rules. This differs from human beings who don’t really operate according to a finite set of rules. In any case, some very interesting complex phenomena can be modeled; and it can be done with very little math. If you have a grid of cells that can take on different states you can watch them evolve over time and observe the various aggregate states that emerge as a consequence of the simple rules governing the individual cells; it’s very Hayekian. To model the FRB instance though would be very difficult. At the very least it would require a lot of thought and I’m not sure whether it wou capture the essence of the controversy.
What would be fruitful is to use multiple online reality worlds like True Life and run them simultaneously using different banking laws. If done right I think it could provide information on something that isn’t necessarily a priori resolvable.
Thanks for the suggestion. But actually I am already very familiar with those kind of things - I am a professional artificial intelligence programmer. My question was about whether simulations were being done by anyone for the sake of economics research.
Humans are (biological) computers - so actually they do.
I agree 100%.
Yes, and chaotic systems are in the end the result of simple cause and effect. However the overall system itself is so complex it’s impossible to realy get a hold of it. If you are a determinist then yes, all human action eventually boils down to some chemical reactions. As all weather is the result of the physical interactions of systems of moisture and velocities. Doesn’t change or negate the butterfly effect. It’s not that human beings are magical. It’s that the necessary access to information and raw processing power simply don’t exist to properly model them. That’s not to say you couldn’t learn something interesting through trying.
Yes and no. I agree that all sorts of assumptions and simplifications, but that doesn’t necessarily mean that it will all be a giant self fulfilling prophesy. I would bet a pound to a penny that the first time the simulation is run the numbers that come out would look very unlike normal economic behavior. My bet is that the size and or frequency of economic cycles would be obviously wrong or funny things would happen like all the wealth would quickly end up in the hands of just a single person or a single bank and everyone else would starve… or that kind of thing. I’m sure it would be very hard to tweak all the parts of the system to get it to behave like real life, and I bet that the authors would learn a huge amount in the process of determining what needed to be changed in order to get sensible behavior.
Computers have no problem simulating randomness.
Yes, but the question is can you learn anything from an approximate simulation. All sorts of areas of science rely on approximate simulations of things the scientist know they can’t simulate in detail.
What size would be ‘wrong’? Define wrong as it relates to the business cycle?
You consider that funny, other people would consider that right and inevitable, and a confirmation of their views.
Random and uncertain are two completely different things. And to my knowledge computers can’t generate random numbers, there is always an underlying formula from which the number is derrived and thus, can be predicted if you have access.
Those sciences study the physical world of inanimate objects where constant, predictable relationships are the rule. We can predict how physical items will react with precision even at relativistic energy levels because their relationships are always a function of fixed measurable characteristics. That is simply not the case with humans. The formulas used to launch a pod into space are not perfect replicas of the rocket, the pod, the person in it, etc., but they are close enough to make the endeavor worthwhile and productive. There is no way, at least to my knowledge, to get ‘close enough’ for a similar measure/model of human behavior and subjective value. Even if the model approximates reality to an extent, it is because you forced it to and that is putting the cart before the horse in this case. Markets don’t behve the way they do because some governor is at the helm making sure certain rules are followed, their behavior is a by product of tons of individual actions. It’s a unique order in the sense that it is being continually destroyed and recreated. In a computer there can be no such destruction and recreation because everything is the result of a pre determined set of behaviors.
When I say “wrong”, I mean “unlike observed business cycles”. So for example say the model had no business cycle at all then that would suggest something in the model was wrong. Or perhaps the business cycle was a perfectly regular sine wave - that would clearly be wrong too.
But what if it happened by (simulated) lunchtime of the first day of the simulation? That would tell you that something was wrong with the simulation. Or perhaps the simulation would lead to everyone ending up with the same amount of money - that would indicate a flaw in the simulation too.
The simulation could fail in a thousand different interesting ways, and resolving these flaws may be very enlightening.
I do not believe there is any need to distinguish between the two for the sake of the experiment. What type of phenomena could show up in a system with fundamental unknowability that wouldn’t show up in a system which merely used pseudo-random numbers?
Not true. I shared an office with a PhD student who was studying artificial populations of “predictors and prey”. The simulations of these showed chaotic behavior with similar characteristics to those observed in nature.
I addressed that exact point in an earlier post.
Computers use a Random Number Generator to give “random numbers”. Some are more random than others, depending on how you design it. It begins with an arbitrary equation that involves more than one variable; one of these is your “random number”. The other variables can be filled in by taking samples from unpredictable/volatile sources. Because most computers have all kinds of processes and analog devices always running they can draw from many sources; for example, the noise in the audio output channel. If you want the result to be limited to a range, such as producing a percentage, you can doing something like use only two digits from the result. It’s only as random (read: unpredictable) as what you feed into it.