Skip to content

Who needs a macroeconomic model?

November 4, 2011

All this recent IS/LM debate got me thinking, people don’t just randomly model things, they have a purpose for their model. Who will use the model and towards what end? There are two basic groups who use economics – the economic engineers and the economic scientists. The engineers will using the model improve policy design and policy implementation. The scientists use the models to explain the world in simple ways to highlight the relevant explanations for phenomena.

The Engineers
The macroeconomic engineers don’t need a macroeconomic model; they can just print money until they hit their target, whatever it is, and then stop. As Mankiw points out, “Recent developments in business cycle theory, promulgated by both new classicals and new Keynesians, have had close to zero impact on practical policymaking.” The total disconnect between Ben Bernanke’s academic work and his actions as Fed chairman confirm that sentiment loud and clear. Modern central banks do not seriously consider rational expectations, futures markets, balance sheet analysis, level targeting or any of the other pillars of modern mainstream macroeconomics. The Fed doesn’t even have an explicit target. Monetary policy, as practiced, is both crude and arbitrary.

The Scientists
For the economic scientists, empirical observation is all that is necessary to describe what is going on in terms of collecting stylized facts. Additionally, since macroeconomic models are mostly disconnected from models of individual behavior, they can’t do any more than simply be a way of framing the stylized facts we have collected. Most economists would carve out an exception for rational expectations models, as they are based on a model of individual behavior. However, Rat Ex relies on ignoring all of the informational problems inherent to the macroeconomy and massively simplifying behavior to the point where the model can be said to be based on individual behavior, but not individual human behavior. It’s based on the behavior of a bizarre species of godlike emotionless identical creatures.

There is one group could claim to need a macroeconomic model – those who decide what the target is in the first place. While hiting the target does not require a model, determining which target to shoot for has profound implications for macroeconomic stability. George Selgin’s analysis of the history of central banking suggests that the primary reason for central banking is to accommodate the spending habits of the government. When macroeconomic goals are considered, they basically boil down to “let’s not let prices change too quickly, because voters don’t like that”. Even with my prefered target, NGDP level targetting, the primary theoretical justification for that is sticky wages, which requires no fancy math to explain and has been around since Hume.

Feedback and Information Problems
Current macroeconomic models have no predictive power, no falsifiability, and no rhetorical impact. Thanks to arbitrage, a model cannot be predictive because any price change it predicts will happen instantly. If a model predicts a stock will drop by 20% tomorrow, the stock will immediately be shorted by the modelers and drop immediately. Markets are information aggregators and credible models are part of the information aggregaged.

Macroeconomic models are unfalsifiable. People have experienced nearly every combination of high/low interest rates, inflation rates, unemployment and GDP growth out there. Every model which simply says when X happens, Y will happen can be proved false by some empirical example. When dealing with AS/AD and IS/LM models, economists can’t even agree on which direction the curves slope, let alone what the implications of the model are. With hundreds of variables which are commonly modeled, whenever a model is shown to be wrong, it is not rejected, but rather slightly modified with the new data and thrown back into the fray.

Finally, models aren’t even rhetorically useful. When the unemployment estimates from the stimulus package were wildly innaccurate (below), exactly zero people on the left were convinced that the stimulus failed and no one on the right was convinced that the stimulus worked and that “the recession had just been worse than we initially thought”, to paraphrase Russ Roberts. It is exceedingly rare for the economic profession to be convinced of a particular theory based on a fancy model of a phenomenon which is not already clear from just looking at the data.

No one believes this:

Conclusion
Perhaps the world is too complex for a decent model to capture the macroeconomy. Economics frequently confronts the existence of feedback loops, self reinforcing cycles and other phenomena from chaos and complexity theories. Since the object of the modeling, human behavior, is itself impossible to model with current technology, how can we seriously expect to model the outcomes of the interactions of billions of humans? Even for highly simple models of the economy, such as IS/LM, have hundreds of variants and policy perscriptions associated with them. Perhaps one day someone will come up with a good model for the macroeconomy, but for now, it seems like a futile and unnecessary task.

Update: Sumner gives a good example.

2 Comments leave one →
  1. noiselull permalink
    November 4, 2011 8:18 pm

    Have you looked at some of Roger Garrison’s work? He does some pretty interesting stuff on macro.

    • November 4, 2011 10:24 pm

      Nope, but I’ll check it out. Do you have any suggestions for where to start?

      He’s the guy who did these slides, right? I’ve seen those before.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: