There has been an intermittent dialogue taking place in response to my previous post on The Economic Modeling of Religion, in which Bob Ekelund, one of the authors of The Marketplace of Christianity and a professor of Economics of Auburn U responded to misgivings about the applicability of the economic model. Recently David George - a colleague at La Salle known for his work in meta-preferences (see his Preference Pollution: How Markets Create the Desires we Dislike) has added a provocative comment to the mix.

So I am starting a new thread here with a top-level post for two reasons: to make sure that newer viewers are award of the dialogue, and particularly Bob Ekelund’s responses and defense of the economic model of religion, and to answer David George’s point concerning understanding and prediction, which will give me more of chance to discuss this continuum.

An excerpt from D. George’s response:

Second, I must disagree with both of you that, to quote Bob Ekelund, "any model must have predictive power." As you, Rich, appear to point out in an earlier entry, "understanding" is also of great importance. Work that I have done tends to focus on exactly such understand[sic] without claiming to be able to predict. The late Milton Friedman brought economics to its present sorry methodological state by asserting in an early 1950s article that "assumptions don't matter." I have seen first-hand the mischief that this can cause. Beginning, I believe, with Thaler and Sheffrin around 1980, attempts to explain internal conflict have begun with the assumption of "two-selves" (or "multiple-selves") residing within each individual. If pressed, advocates of these models will probably stress that person isn’t really “two-selves” but simply behaves "as if" he is. As I have argued extensively, this does little to further our understanding of internal conflict. To tell someone trying to understand her internal conflict that she has more than one self begs the question, to put it mildly. To explain that the preference that moves them to act is not the preference that they prefer having, in contrast, "sheds light." To be in the grip of crummy preferences is to rationally act and to still be respected as a "whole." In addition, it permits normative evaluation of market shortcomings (markets are ill equipped to create preferences that we prefer having) while at the same time not particularly suited for making predictions. The two-selves models, in contrast, do not have the same normative weight. My point: assumptions matter if we are to gain "understanding."

OK, I totally agree that assumptions are crucial to a true understanding of any phenomena via modeling. But, and this was my point in my original post about the economic modeling of religion, if there is no predictive power, I question what the model provides other than an explanation for why things are the way they are. I’ll even go so far as to say that, if there is no prediction, there is no understanding.

Now, with apologies to D. George, this is a pretty harsh view. I am lead to it, however, by wondering where the belief that the model provides understanding comes from if not from some "test" of the model via prediction. Maybe it’s the word model that has me going here. A model that provides understanding ONLY is really only an explanation of the current situation - one that might "sound right," but which is lacking any inductive way of measuring its merit as compared to another "model."

There’s an interesting connection to Chaos Theory here. Because of sensitive dependence on initial conditions that is the hallmark of many a non-linear dynamical system, it would appear that one could have complete understanding of a system (i.e. the equations that determine the interactions among the system members are know to be exact) but no predictive power. So doesn’t this blow away my argument about the need for prediction? How can Chaos Theory be used to model anything at all if non-predictability is built in? The difference is that Chaos Theory, in a nicely recursive way, predicts its own unpredictability. Not only that, but because of the constrained randomness that is the best way of describing the output of many a chaotic system, it does so pretty accurately.

So I challenge anyone to come up with a model that only yields understanding, but no prediction.

I predict that, in all cases the "model" is better described as an explanation of the current state.