Mittwoch, 25. April 2012

Abrechnung mit RBC-Modellen

Auf Noahpinion steht seit vorgestern eine Abrechnung mit RBC-Modellen (Real Business Cycle Modellen), die sehr lesenswert - um nicht zu sagen amüsant - ist.Noah Smith zeigt darin nochmal anschaulich dar, in welchen Punkten diese Modelle (die lange Zeit die Modelle der Makroökonomie waren) an der Realität vorbeigehen.
"It has often been said of the Holy Roman Empire that it was "neither Holy, nor Roman, nor an Empire." However, that joke has gotten a bit stale since Voltaire wrote it in the 1700s, so I think it's time for a new one. Real Business Cycle models, it turns out, are neither Real, nor about Business, nor about Cycles."
Dann ist es auch kein Wunder, dass diese Modelle schlechte Prognosen liefern:
"Basically, here's how an RBC model works. You take every factor of production you can measure - capital, labor, inventories, etc. - and you factor it out, and then you're left with the part of production you can't explain, which is called the "residual". You then label this residual "technology", and you assume that it moves according to some sort of simple stochastic process - for example, an AR(1). The rest of the model is just a description of the ways in which the rest of the economy responds to that AR(1) technology "process". In RBC models, this response is usually as simple and uninteresting as possible; pretty much everything is driven by the uber-simplistic movement of "technology".
In other words, if I want to make a forecast using an RBC model, that forecast will be based on the assumption about tomorrow's level of "technology" - i.e. the part of the model that doesn't come from data we can directly measure - and that level, in turn, will be "predicted" by nothing more than the simplest stochastic equation imaginable!"

1 Kommentar:

Ulrich Fritsche hat gesagt…

Hier meine Antwort im Original-Post:

Nice piece. I am not a fan of RBC models but you have to be more precise. Three points.

@Cycles: Since the papers of Frisch, Slutzky and Samuelson, "shocks" are modelled as kind of "fundamental" drivers of an economy which can be described using the comparison of a rocking chair. The economy tends to market clearing and equilibrium and then some forces drive you away from that point and the chair oscillates. Impulse and propagation. Therefore cycles -- boom and busts -- are not only driven by the shocks but by the structure of the economy. In the case of Samuelsons famous multiplier-accelator model it was the interaction between households and businesses which defines the propagation mechanism. So, even in simple RBC models we can have (quite regular) cycles.

@ Businesses and Real: I completely agree. New Keynesian DSGE models added monopolistic competition and more detailed firm behaviour and a lot of channels to make the impact of demand shocks more prominent. However what is more a kind of problem in this models is the Frisch-Slutzky-Samuelson paradigm of an inherently stable market clearing process.

@ Forecasting: This is a misunderstunding on the role of models in thinking about the world works. We build models to reduce complexity and focus on the central channels. We do not build models to forecast (you never try simple forecasting techniques with sophisticated models....)

Kommentar veröffentlichen