Tuesday, February 7, 2012

Wow - a Stephen Williamson post I agree with 100%!

Williamson's blog has been disappointing - I used his textbook as an undergrad and liked it a lot, but his blog has mostly been mudslinging. I almost took him off my blogroll last week, in fact. But I'm glad I didn't this is good:

"Why Economists Are Right
Here's an
excellent piece by David Levine in the Huffington Post. There are some related ideas in this article of mine. Excerpting from the latter:

But prediction need not always be the criterion for success of an economic model. Clearly, if we are judging a forecasting model, we want it to predict well, in some well-defined sense. But in other cases forecasting is not the name of the game. In arbitrage pricing, under some assumptions the model implies that changes in asset prices cannot be successfully forecast. By the criterion of prediction, the model is indeed a total failure. It tells you that a monkey could do as well at predicting asset prices as an economist who understands the model. Yet the model is actually not a failure, as it teaches us something interesting.


With financial crises, a similar issue arises. By its nature, a financial crisis is an unpredictable event. We could have an excellent model of a financial crisis. The people living in the model world where the financial crisis can happen know it can happen, but they can't predict it, otherwise they could profit in advance from that prediction. Similarly, an economist armed with the model will not be able to predict a crisis in the real world. A nice example is in Ennis and Keister (2010), which is a variant of a Diamond-Dybvig (1983) banking model. In Ennis and Keister’s world, a rational and benevolent policymaker and a set of rational consumers live in an environment where a banking panic can happen. When the policymaker sees the beginnings of a panic, he or she starts to intervene, but rationally discounts the severity of the panic until the panic is too full-blown to actually prevent. Thus, the panic can happen even if the policymaker has the right model, and the policymaker with the right model indeed cannot predict the panic."

Now - how do I get other economists I clash with on a regular basis to understand this?

6 comments:

  1. 'With financial crises, a similar issue arises. By its nature, a financial crisis is an unpredictable event. We could have an excellent model of a financial crisis. The people living in the model world where the financial crisis can happen know it can happen, but they can't predict it, otherwise they could profit in advance from that prediction.'

    Isn't this assertion quite blatantly contradicted by Steve Keen?

    ReplyDelete
    Replies
    1. Explain.

      I think lots of people saw the prospect of instabilities. You can observe those prospects for instabilities in a model of a crisis too.

      What's unpredictable is how the instabilities play out. Many people saw a financial crisis coming. To a certain extent I guess you can call that a "prediction", because the word is vague. But it's not really a forecast as traditionally understood.

      And regardless - the ability of a model to forecast the future is not a prerequisite for a good model of a complex system.

      Delete
    2. how the instabilities play out...

      we don't know exactly what we happen when a car goes around a curve at too high of speed, but we still have speed limits.

      and, here Williamson just flat out lies:

      "otherwise they could profit in advance from that prediction" This is just absolutely false. It assumes that some investment vehicle exists by which to make a profit. Lewis had a great book last year, The Big Short, entirely focused on the fact that people who could foresee what was going to happen had to go to extraordinary lengths to profit.

      Insofar as I know, there is no way to short housing prices. You can infer that housuing prices are going to go down and therefore something else will move in parallel, but there is no way to short housing.

      Delete
  2. Daniel Kuehn, it appears that Dr. Brady may be correct that financial crises are ergodic and predictable, according to this article I found over the web.

    http://iopscience.iop.org/0295-5075/90/3/30004/fulltext/

    Dr. Brady also writes about the ergodic/non-ergodic distinction that Paul Davidson gets confused about in the following review on Amazon.com.

    http://www.amazon.com/review/R32PPK2MQ5SQUG

    To sum up, I side with Unlearningecon that financial crises are perfectly predictable, even if we may not have exact precision as to the predictions.

    ReplyDelete
  3. I think that's based on a misunderstanding of what "prediction" means in this context. I always understood "prediction" in social science as meaning that one could empirically verify the model. In other words that the model would predict parts of the world not yet seen. Of course, since the entire future is "part of the world not yet seen", it is a very attractive place to look for empirical verification, but it need not be the only part. As long as you are testing your model against data which was not used for the creation of the model, you're "predicting".

    ReplyDelete
  4. Marx had a perfectly operable micro/macro model of capitalism which predicts systemic crisis. When these occur is not predictable leading to marxist economists predicting 5 out of the last 3 recessions. Unfortunately for ideological and historical reaons this line of thought has suffered from arrested development vis a vis neo-classical.

    ReplyDelete

All anonymous comments will be deleted. Consistent pseudonyms are fine.