24 March 2011


Someday, scientists will learn that mathematical models only predict what they are constructed to predict. At that point, they will have finally intellectually evolved to the point of a 13 year-old computer programmer. Because computer models are neither studies nor are they science.

For some reason, science fetishists routinely forget that all scientific disciplines are inherently axiomatic, and that whatever truths science reveals are inherently subjective in nature.  Because of this, said fetishists tend to accept mathematical models as proof of argument, when in reality the only thing mathematical models prove are that computers are capable of logic.

It is important to remember that in order for a conclusion to be correct, the logic must be correct and the premises must be sound.  As such, if a mathematical model is proved incorrect, then it is obvious that the assumptions upon which the model the model is based are inherently flawed.  The lesson to take away from all this is that one must always question the underlying assumptions.  This is especially true in the computer age, for the logic of computer-generated mathematical models is virtually always sound.

Incidentally, this addresses a fallacy found in John Case’s post from yesterday, specifically in reference to his second “point:”
Austrian economists also argue that mathematical models and statistics are an unreliable means of analyzing and testing economic theory, and advocate deriving economic theory logically from "basic principles" – read "divinely inspired principles" – of human action.

Again, the reason why Austrians reject mathematical models is because they are inherently axiomatic.  This is why CPI and GDP are such laughable measures.  CPI excludes fuel and food costs, yet purports to measure price inflation.  Seeing as how fuel and food are fairly common purchases for all people in a given economic system, it is absurd on its face to exclude them from a metric that’s supposed to measure systemic inflation.  GDP also has its own problems; for example, government spending is seen as production, not consumption, and is considered to be as efficient in production as private production.  There are other concerns for both metrics, but it should be obvious that accepting mathematical models based on these models is foolish, to say the least.

Furthermore, given the sheer amount of knowledge that one must possess in order to build data-driven models, it is simply easier and more correct to analyze economic phenomena from a “basic principles” standpoint.  Plus, human action is easily observed, and one can generally determine why a given person behaves a certain way.  As such, it is easier to simply base policy on how humans actually behave, instead of basing it on an inherently subjective model that con only say what programmers tell it to say.  All the math in the world cannot change human nature.


  1. "Plus, human action is easily observed, and one can generally determine why a given person behaves a certain way."

    Don't agree. The 'why' is exactly what an Austrian is not interested in. 'Why' can not be known, and if it could, couldn't be generalized for anyone else than the acting person. In fact, it can't even be generalized for that one person.

    Also, your sentence preceding the quoted one does not make sense. Sheer amount of knowledge does not help you to build better data-driven model. What's inherently wrong with the metric approach to economics is that the pre-definitions and 'axioms' of the models are wrong. It also seems to me that the more knowledge those metric guys use, the less value will have their models.

    And Austrian economics is not simple. Evidence for this is found in the vast amount of economists who don't understand it. Well, maybe that's not evidence per se. It could also mean that stupid people get attracted to complex things and smart people to simple ones. Who knows.

  2. strange. do anonymous comments get deleted around here?

  3. That is the great failure of positivism. This type of modeling works great with matter -- physics, chemistry, even biology.

    But when applied to social science, it fails miserably. The variables are too many and our behavior is to complex to even model. The Austrian argument of the calculation problem can be generalized and applied to essentially every human endeavor.

  4. @artie- no. they apparently get caught in the spam filter.

    "Don't agree. The 'why' is exactly what an Austrian is not interested in. 'Why' can not be known, and if it could, couldn't be generalized for anyone else than the acting person."

    I probably wasn't being clear enough here, but I was referring to Misesian rationality, wherein a person's motivations are inferred by their behavior, making all people, by definition, rational creatures. (I've discussed this before in reference to behavioral economics.) Essentially, this is a tautology.

    "Sheer amount of knowledge does not help you to build better data-driven model."

    It would if we could account for all the relevant variables, and attach the proper weighting to each. Some models fail because they focus on the wrong things, others fail because they don't focus on enough tings, etc. False assumptions an certainly be a problem when building models. Some falsity is the result of human stupidity, some falsity is due to human ignorance (i.e. there are some variables we cannot even conceive of).

    "And Austrian economics is not simple."

    Personally, I've found Austrian economics to be quite simple. Elegantly so, I might add. My limited experience is that unintelligent people tend to not understand economics at all, super-intelligent people tend to identify with neo-classicists or Austrians, and the moderately intelligent identify with Marxism and/or Keynesianism. I think the reason why the moderately intelligent get caught up in the left-wing camp is due to the technicality of the language (and models) used, which makes it seem more intelligent than it really is. It's the people who are just shy of possessing a genius-level intellect that most greatly underestimate their stupidity and ignorance, at least in my experience.

  5. @artie- also, it's impossible to build a predictive model without ever asking why. I believe that's the reason why praxeology exists. Knowing that humans act is redundant.

    Now, we can never know with one hundred percent certainty that someone will behave a certain way in light of a specific set of variables, but we can predict what most will generally do in light of a certain set of variables. Some arguments against, say, corporate welfare are based on a general understanding of human motivation.