Macroeconomic Computer Modeling?

Economics Economics Questions
21

You said nothing that fundamentally contradicts them, yet then you come out with this silly pronouncement, relegating the works of men like Hoppe, Mises and Hollis to the status of “ramblings” because you do not understand them. I’m not a Kantian so I do not agree with how Mises phrased/argued his position, for the most part, and I think Austrianism has an “empirical” element in the sense Rothbard conceived of it, namely in its concept-formation and retroactive control over its axioms (as Smith puts it), but that is nowhere near endorsing the hypothetico-deductive methodology or abandoning Mises. “Dogmatic” my ass…

-Jon

22

Jon,

Given the details and outcomes of the Game Theory article I linked to earlier, I think Game Theory is breaking out of being a conceptual tool and into something far more useful - a working, predictive modeling framework. I think the central area of breakthrough is around being able to express in math human nature within a definable context. In the article, the scientist in question was able to model and predict with accuracy the political machinations of more than one country as to who would emerge the eventual leader - this is as human as it gets and as far away from conceptual and into actual as possible. I feel the framework of axiomatic-deduction has applicability to the new style of Game Theory for modeling the frequency and amplitude of future business cycles. However, sadly, I lack the math skills to take this any further myself.

23

Yes, it could be used that way.

-Jon

24

A great idea.

25

Among the numerous variables the model would have to take into account, evil and the potential for human beings to abuse any econonmic model would have to be factored in. In addition, the potential for human beings to improperly implement the superior model for the given situation would have to be factored in. Models will react differently to certain levels of abuse and mismanagement (more or less forgiving). It would be great if such a model would produce two results:

  1. The model with the highest positive elasticity.

  2. The most inelastic model.

26

As a comp-sci grad myself, I considered getting into something like this.

However, I have serious doubts about it’s feasibility. Modelling human behaviour is impossible for a computer to do, let alone doing it for a global population. At best, you might be able to model behaviour in terms of odds and percentages but the more you generalize behaviour, the less accurate your results will be. Your results may approach uselessness pretty quick.

I think economics has a kind of butterfly effect factor involved. One purchase here could cause economic collapse elsewhere. You may not be able to predict economic conditions any better than you can predict weather conditions.

I think this is what draws me more to Austrian economics over anything else I’ve heard. Austrians seem to draw from historical precedent above anything else. And when dealing with something as volatile as economics, I have to put my trust in history more than I trust singular mathematical predictions.

27

Coming from an engineering background, I thought of plasma physics. The billion-dollar 40 year history of failure to achieve fusion’s plagued by the same difficulties as modeling economic systems.

Just doing a few hours of research, with my economic knowledge limited to a couple basic Austrian theory books, I found the method I had in mind already in use:“Agent-Based Computational Economics” (ACE) http://en.wikipedia.org/wiki/Agent-Based_Computational_Economics

ACE is based on relations of programming agent-objects’ methods and attributes, like the popular Cellular-Automata “Game of Life”. I see there are simple examples of trader-agents which would need much more elaborate, personality and emotional detail to model reality. And since real people’s do myriad unrelated events (selling a pet-food stock because a dog bite, etc). A predictive model would need immense trivial detail. And as in chaos, there’s emergence of markedly new dynamics at increasing scales of social order (e.g. family, village, state, nation, world). In plasma-physics the particle-in-cell approach is assumed to practically realize an approximation.

Perhaps its possible to technically depict, if not predict, as others have noted here, specific events by selecting specific parameters at each scale of organization?

At the risk of changing the subject, this model could depict the Austrian approach and the error of trusting central-planners to the public, as the Hake’s simplistic Return-To-Serfdom cartoon once was, but now using the latest video-game 3D rendering (something like the “Nightmare Before Christmas”, etc). It would give students a fun video-game model to play with as scenarios are re-enacted with changing parameters. It would be a game like the popular “Civilization”, but with useful economic and historical realism.

There’s a website, Sourceforge, where such a project could be started (if its not going on already). I suspect the theoretic talent I certainly lack is here. There’s a large community of video-game modders that could hopefully be tapped for the artistic talent to add the artwork to already-existing games. I suspect the technical time and effort required for fine-tuning the model would scare off any commercial, if not academic interests, however.

Suggestions?

28

I think it would be interesting to create a virtual world with a limited supply commodity/money built-in, and then just let people join and go nuts. Maybe something like Second Life, but of course with a “natural” hard money available instead of a fiat currency (although in this virtual world I would assume users would be able to create a fiat currency if they’d like).

Maybe such a virtual world could be built on one of these:

Croquet - http://www.opencroquet.org/index.php/Main_Page (open source)

Multiverse - http://www.multiverse.net/index.html (not sure if it’s open source)

The Interreality Project - http://interreality.org/ (open source)

This would be far different than simulating an economy, though. You’d have a virtual world with real people acting in it, and real wealth being created and traded, even between the virtual world and the real world. Any intervention would be harming real people and their real wealth. Once real world wealth gets involved, though, you can be sure real world governments will want to interfere.

29

That many “Austrians” reject modelling I think is a shame.

Modeling is a crucial aspect in economics. Take RBCT, for example. Furthermore, if you want to do almost anything outside of political economy - for which you’re probably going to have to work for the government - then you need to know your way around a financial model.

Without a basic understanding of computer models, you won’t get anywhere in the real world.

30

I may be wrong (probably am) but I don’t think spending, or for that matter sites like 2nd Life would be very popular. If I had family, it would scare me to think they were practicing spending money [6]. I’d rather they be shooting something else off.

That wasn’t what I was getting at, I was talking about simulating a real economy, just being able to use 3d icons at each level of scale -

  • ships, planes and dollars flying on a globe between countries
  • trucks , planes, trains and dollar within countries
  • …
  • little people running in and out between houses, stores, hospitals with dollars
  • surging floods of dollars between banks, the fed, and the treasury

Then have a little control panel to vary parameters to show how things like interest rates, reserve fraction, et. affects the economy. Typing numbers in a keyboard and watching spreadsheets isn’t nearly as “fun” as watching an artistic 3d rendering. Especially if you want to make a documentary to show your stupid neighbors what they are enabling their government to do to them!

Thinking more about it though, to be realistic and simulate historical events could be too computationally intensive. You have several sorts of citizen, several sorts of families, several economic sectors, etc. to build massive databases of demographic data for, then recurse back from the macro to the micro scales to reiterate the myriad permutations of parameters. Plasma fusion physics is much simpler [:)]. Perhaps when quantum computing matures.

BTW, I meant to say Hayek’s “Road to Serfdom” in my initial post.

31

Then you need to read Physician-Mathematician-Computer Scientist Stephen Wolfram’s treatise on science, because he shows exactly why most physical phenomena cannot be expressed through mathematics.

Shortly said, there is only a very small part of the universe that can be modelled with mathematics.

32

Perhaps because it is irrelevant to pure economics. I don’t disagree that modelling has its uses (such as putting the ABCT in terms modern economists better understand), but Austrians have sound reasons for rejecting the neoclassical paradigm, which confuses “modelling” for sound theorizing (which is more the fault of its blind obedience to the hypothetico-deductive methodology than anything else.) That said, maybe Austrians with sound theory behind them could put the models to better, more correct uses.

-Jon

33

There were these 2 links I found from your link that had a lot of resources to start learning more:

Society of Computational Economics — http://comp-econ.org/

Agent-Based computational Economics — http://www.econ.iastate.edu/tesfatsi/ace.htm

I agree that such an Austrian modeling game would be fantastic to help convince others, or at least to help communicate theory in a way that is more accessible to people. I also tend to be the “don’t tell me, but show me” type when it comes to learning something new.

Developing an open source project in SourceForge is a good idea too, for the graphics and gaming environment development, wouldn’t be difficult to seek volunteers. Now if only somebody like Bruce Bueno de Mesquita was willing to design and test the models according to Austrian theory . . . Hmmm . . .

34

Would you mind expanding on this a bit more or pointing me towards something that does? Perhaps something in Rothbard’s writings that shows how he’s different from Mises?

35

I followed several of the links at http://www.econ.iastate.edu/tesfatsi/ace.htm before I posted here. After reading a decument about ACE from there, I did a search here at mises.org to see if anything was hear about modeling and found this thread.

Econ is a new interest for me, an offshoot of politics, having just finished Ron Paul’s “Revolution” reading list and gotten “What government has done to our money”, and indeed in great part, having my representatives do that 700 Billion bailout after so many of us protested, then watching the system melt down.

Oh the joy of repeatedly watching Bush, Paulson and Bernanke repeatedly attempt to reassure us, then seeing the market drop hundreds of points! THAT’S MORAL HAZZARD[8o|]!!!

Any why any modeling program must account for zeitgeist, greed, corruption, honesty, altruism, patriotism, disaffection, indifference in the economy, its actors and influences.

But I digress. I suspect before too long, there will be lots of students and professionals with a lot of time on there hands, asking “what happened to us? [:'(]”, finding places like this, finding out the truth, and available for action!

36

Rothbard wrote far too little on this, but I would direct you in the direction of Geoffrey Plauche’s paper on Aristotelianism and apriorism and Long’s paper on non-precisive abstraction (I forget its title) and Barry Smith’s writings on the LVMI site.

-Jon

37

Thanks for the help.

38

This topic has always fascinated me. I think building accurate models could be very important in promoting freedom, not to mention the potential money you could make in the marketplace.

I’ll have to check out some of the links that were posted on this thread, but I think it’s nearly impossible to effectively model the economy as a whole right now. I just don’t think the technology’s there to model the human thought process to the extent that would be needed for an all-encompassing economic model. You could create a model based on various statistical measures but I think that approach is fundamentally flawed if you believe that the mechanics of individuals making exchanges is the essence of the market. I could write a very reasonable looking program right now that could “prove” central planning works based on economies of scale, reduced customer acquisition cost, etc.

I think models that are very narrow in scope might be the way to go; ones that demonstrate a specific aspect of our overall system. These “mini” models would still be based on individual actors, but they might only have a single decision point. This would make the model much easier to build and understand. One example I can think of was in Dawkins’ “Selfish Gene”. It’s been a while since I read it but, in short, the actors could choose to screw each other over or cooperate and there were different payouts for each combination of actions. If you’re familiar with the prisoner’s dilemma, it’s basically a version of that, but repeated many times. The model showed that aggression doesn’t pay in the long run, which is a good demonstration of the non-aggression principle.

Maybe after many mini models are demonstrated as being accurate, they can be integrated into larger models, and then those larger models integrated with each other and so on. Advancements in technology should also help move this process along.

39

It’s always nice to see my intuitions confirmed by an expert.

40

True “AI” in computer models has not yet been achieved in the computer sciences. Most computer models are based on historical data. As we all know the historical view point on the the success or failure of economic history can very greatly from historian to historian. For this reason a true and accurate computer model would not be possible at this time. We can not get intelligent economists to agree on whether the Keynesian economics that FDR implemented to get us out of the Great Depression help of hurt our economy. This does not even take human greed into consideration. For this reason I do not feel an ubiased computer model would be possible.