Monday, September 28, 2009

When does the model NOT work?

In the wake of the Great Recession, there has been a growing cacophony of calls to do something about macro-economics theory. There is a growing realization that economic models have a lot of simplifying assumptions that we have ignored or forgotten about when forming economic policy. What Paul De Grauwe wrote is just one example out of many:
How to resolve this crisis in macro-economics? The field must be revamped fundamentally. Some of its shortcomings are obvious. Before the financial crisis, most macroeconomists were blinded by the idea that efficient markets would take care of themselves. They did not bother to put financial markets and the banking sector into their models. This is a major flaw.

There is a deeper problem, though, that will be more difficult to resolve. This is the underlying paradigm of macroeconomic models. Mainstream models take the view that economic agents are superbly informed and understand the deep complexities of the world. In the jargon, they have “rational expectations”. Not only that. Since they all understand the same “truth”, they all act in the same way. Thus modelling the behaviour of just one agent (the “representative” consumer and the “representative” producer) is all one has to do to fully describe the intricacies of the world. Rarely has such a ludicrous idea been taken so seriously by so many academics…

The basic error of modern macro-economics is the belief that the economy is simply the sum of microeconomic decisions of rational agents. But the economy is more than that. The interactions of these decisions create collective movements that are not visible at the micro level.

Understand your assumptions
What a lot of modelers have forgotten is the first rule about model building: understand your assumptions.

Under what circumstances will the model fail?

If you can’t answer that question, or if you believe that the model works under all conditions, then you haven’t fully thought through the modeling process.

Inflation-deflation timer
Take the example of my inflation-deflation timer. While the backtest results look promising, when does this asset allocation model fail?

The inflation-deflation timer was built for a bifurcated environment where investor sentiment is likely to move from one extreme of inflationary heat to the other extreme of deflationary freeze, but not much in between. If other factors were to emerge in the years to come that drives growth patterns, e.g. nanotechnology becomes a transformative technology or the emergence of longer human lifespans changes investment and consumption patterns, the performance of the inflation-deflation timer would become unstable.

I try to think about the basic underlying assumptions behind my models. Do macro-economists do the same? Do you?

Model update: The inflation-deflation timer continues to signal a "buy inflation" reading.


Patrick said...

Since we´re on the topic, how likely do you see the emergence of nanotechnology and life extension as being? More interestingly, do you think we´ll still be using fiat currency by then?

Cam Hui, CFA said...


I have no idea how long nanotechnology or life extension technologies will become a real driver in economic growth. My best guess is that they are at least 5-10 years off.