How economists are wrong to underplay computational simulations

Simulations are taken very seriously in all sciences but economics. This great article in the New York Times by Mark Buchanan makes an excellent case in favor of doing more simulations to deal with thorny problems in economics. Thanks to Rob Gilles for sending me the link. I completely agree with Buchanan. I am slowly learning how to do simulations, and am amazed at how good they can be at proviging insights into truly complicated problems, including the functioning and, lately, malfunctioning of markets.

Advertisements

3 thoughts on “How economists are wrong to underplay computational simulations”

  1. I agree computational models gives us a powerful tool. The aid of a computer frees us from many of the simplifying assumptions we need to make to gain tractability on an analytical problem. However I don’t see the incompatibility between the traditional economic theories and theoretical approach and the computational approach. I view them as complimentary. I have noticed a tendency among agent-based modelers to presume that they are studying or developing a new economic non-equilibrium theory. I do not agree. I think they are studying a different part of the problem. The computational approach allows us insight (“behind the curtain look”) at the transition dynamics that take place in an economy. Traditional equilibrium theory studies the end-points of those transisitions. It gives us insight into the general direction but not all of the intermediary effects.

    Because traditional general equilibrium theory can not always anticipate all the ways the transition process might get redirected as information changes or other shocks enter the system, people are quick to criticize equilibrium theory as irrelevant (the system is never in equilibrium, etc.) I think this is a mis-perception among (mostly non-economists) about what general equilibrium theory is and what questions it is used as a tool to answer.

    I think economists need to use both approaches. One informs the other. ABM researchers, for example, say their agents have simple decision making processes that use limited information. Theoretical/mathematical economists call this boundedly rational and study it with analytical tools. ABM developers could save themselves a lot of modeling time to read this literature and incorporate it in the programming. Futhermore it would give more theoretical structure to the models and, I think, aid in making the models more robust to small changes in the parameters. (One of the criticisms of these models is that they depend heavily on the researchers built in assumptions about the agents and their “learning” algorithms.)

    These are some preliminary thoughts. I need to study agent-based approaches more.

  2. There is a legitimate place for theories that emphatically deny that the economy, or any social system, ever even approaches equilibrium. Simulations let us look at this question, which equilibrium-based theories cannot even formulate.
    Of course I have a soft spot in my heart for equilibrium, but I’m excited by the possibility of studying what we can without the need of the equilibrium hypothesis.

  3. Even if we never get near the equilibrium (because there are always forces acting on the system; entrepreneurs 😉 for example), there might still be tendencies pulling towards some stabilization. (I think the computational approach calls this a steady state instead of an equilibrium.)

    Do general equilibrium theorists believe the economy ever reaches the static equilibrium? (I was under the impression that most acknowledge it doesn’t but the theory still provides useful benchmarking and perhaps even isolating effects. An agent-based modeler admitted that sometimes it is difficult to know exactly what in the model drove a particular result.) If we know in theory what type of effects we might expect, then we might be able to better analyze the results from the simulation and see why perhaps we got unintended consequences (the meaning of an unintended consequence finds meaning when paired with the intended consequence).

    Again just some thoughts (not well articulated). I’m a neophyte at these approaches and also excited about the research possibilities using agent-based models.

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