I would appreciate someone clarifying whether and how "Econometrics" fits within Austrian economics. I have described Austrian economics as "non-statistical" or "non-mathematical", based on deductive logic from certain basic axioms of human action. The push back to that has been, "Don't you need to test whether your theory works 'in practice' and therefore need the validation of econometric studies?"
Peace and Freedom
And how would one test an economic theory? :)
Relevant articles:
Frank Shostak's "What Is Wrong With Econometrics?":
http://mises.org/daily/940
Bob Murphy's article "Econometrics: A Strange Process":
http://mises.org/daily/1001
Austrians use econometrics as a foreign language that they occassionally have to speak in order to illustrate theory or historical points. You can't really test economic theories because it's impossible to predict in a quantitative manner, and if you do somehow manage to, your prediction becomes a part of the experiment and ends up disproving itself. For example, if I knew the market was going to crash in a year, really knew it somehow, at the very least my behaviors would start to be incorporated into the system and work to change that result. And if people found out I had this knowledge, so would theirs. And the market would crash sooner because of that, making my forecast wrong, even though it was right. Bottom line, you can't run a controlled experiment in this field.
Now it's the NeoClassical argument that the unrealistic models they use are 'close enough' to allow for predictions, and it's the predictions that matter. Austrians disagree and fee that what's left out of the models, and which generally can't be incorporated because of difficulty modeling unpredictable human behavior, makes them one big fallacy. Because the theory that underpins them is secondary to the modeler's desire to predict or forecast, what they end up doing is committing one big correlation for causation fallacy.
The objection is also that it is impossible to construct 'objective' econometrics. Even something as simple as CPI, how do you do it? What goods do you include? What units? What weight do you assign to each? How do you account for quality changes over time? There's no non arbitrary answer to any of those questions. Or say it's price information, that's just historical information. It doesn't tell you anything about what something is actually worth now or will be in the future.
Overall there are no constant relations between the factors that produce 'the economy'. The market is a social phenomena, it results from the interaction of billions of people every minute of every hour of every day. It doesn't exist apart from this interaction, nor can it be predicted, but neither is it random. It is the result of purposeful action. And since nothing you can potentially measure on the market can give you any insight into the why of many people's decisions, much less what they'll be in the future, trying is kind of a lost cause from the begining.
Now it's the NeoClassical argument that the unrealistic models they use are 'close enough' to allow for predictions, and it's the predictions that matter.
No, i'ts not. Mainstream economists use empirical data to see if it matches up with economic theory. If you've ever taken an econometrics course you'd know that they tell you the shortcomings of predictions and how you can poke holes in any econometric study.
So justin, if they are aware of the shortcomings, why they keep doing it? Broken clock tactic?
"No, i'ts not. Mainstream economists use empirical data to see if it matches up with economic theory. If you've ever taken an econometrics course you'd know that they tell you the shortcomings of predictions and how you can poke holes in any econometric study." - justinx0r
Then what is the point of the studies, pray tell? If they don't predict reality and aren't even expected to then the purposes of the models is...?
Needless to say, the motto of the Econometrics Society is "Science is prediction." Wonder why they'd mention prediction in here.
The equations NASA used to send the rovers to Mars weren't necessarily perfect models for the rovers. The weights weren't perfect, the air resistance predicted for the rocket wasn't perfect, etc. But they were good enough to launch the thing and succeed in getting it to another planet. To stealan analogy from Bob Murphy, NeoClassicals tend to view their models the same way. No, they're not perfect. They'd be idiots or lunatics if they made that claim and they know it. However the contention, either explicit or implicit, is that they are good enough for the purposes of prediction. And what's more, if they follow Friedman's extreme positivist approach, then the predictions matter more than the underlying accuracy of the model.
And if they weren't for prediction and that wasn't the point, then what is economic modeling but history's biggest academic circle jerk? You have to move beyond what they tell you in class and start thinking about implications and consequences here.
justinx0r:Mainstream economists use empirical data to see if it matches up with economic theory.
What do they do if the data doesn't match the theory?
My last post here explains that.
Econometrics is far superior to conventional "post hoc ergo propter hoc" nonsense that passes for evidence in forums/blogs/etc. "In 1978 Norway had 60% tax rates and record growth - I guess that means tax rates are good".
Of course there are at least two things wrong with this. One is that we are only identifying correlation, not causation. Did government spending cause high growth - Did high growth cause government spending - Are the two unrelated - Are they in fact inversely correlated and this is just noise? So the debate can go on forever. If you've ever tried to argue evidence, as I have, this crap keeps coming up over and over again, and both sides can just selectively claim that the good prima facie evidence supports their position, while the bad prima facie evidence is actually just bad. Zzz The second problem is that identification of relevant evidence presupposes a theoretical framework through which you can sneak in confirmation bias. For example, if democracy is representation of the people, then by definition all the failed "democracies" in latin america don't count, and only the nice democracies in euroland count. Similarly with "free markets". If a free market ever failed and violated property rights, we would just say that's not a real free market. So I would look at something less semantically bound such as "decentralized legal systems". But of course you can never get this nuance across to your opponents...
The reason that econometrics is superior is because the model builds in causation. The direction of causation isn't clear, but there's a goddamn equal sign there that pins you down. It also allows you to describe non-linear relationships across time. The above methods are implicitly linear, which is never justified. Its just the default way our brains work.
The severe limitations of econometrics are given above. I just don't want anyone thinking that they're fundamentally different from any other empiricism.
The "great" thing about data is that it can be made to fit almost any theory. That's the greatest weakness of empirical economics.
Indeed, it would be interesting if economics was held to the same standard as the natural sciences in actually being required to make positive predictions about future phenomena, and not simply retrodicting past data. As Niels Bohr said:
"When the King is far the people are happy." Chinese proverb
For Alexander Zinoviev and the free market there is a shared delight:
"Where there are problems there is life."
I built an econometric model to troll lefties at debate.org Heres a debate I used it in. What basically ends up happening is they try to argue with it for a bit, then use linear/isotemporal counterarguments, then give up and say you must be biased, at which point you accuse them of being ideologues who just "know they're right". At the very least it lets me throw out empiricism. Its the crutch of lefty economics. "It works in Denmark..." I have a .95 R2 on Denmark punk.
An econometric study could tell you (for example) if, statistically, bank runs increased after the implementation of deposit insurance during the Great Depression (just pulling an example we used in one of my econometric classes). You could use a model to try and predict but that's not their sole purpose. In fact, econometric predictions are torn apart all the time.
justinx0r:An econometric study could tell you (for example) if, statistically, bank runs increased after the implementation of deposit insurance during the Great Depression
Right. And lets say that you had a theory that predicted that bank runs would decrease but the study showed that they increased. Then what?