Modelling in Economics, redux

A recent blog comment by "Cranky Observer" on Brad Delong's blog caught my eye. (The context: Brad's discussion of a model for stock valuation based on economic concepts like 'marginal cost of goods sold')

I am approaching 25 years in manufacturing companies of various types ...I have not ONCE encountered an organization that operates in the hyper-rational, mathematical manner that most posters here seem to believe is the norm in actual business. In fact the most successful of the places I have worked deliberately drove out anyone who attempted to make decisions in that manner in favor of people willing to spend long hours with the customer then make decisions intuitively.

I have also worked extensively in business information systems and associated data warehousing and I can tell you with absolute certainty that 97% of the organizations in the world have no bloody idea what their marginal cost is for purchased paperclips much less the product that they make.

Wow, that's quite an indictment. My response was as follows:

Cranky, when I started my economics education (after years in IT), I too was hostile to the use of mathematical modelling that did not take institutional and information realities into account. However, I am beginning to think some* of my hostility was misplaced for two reasons.

Firstly, simple equations governing the behaviour of stylized corporations are only meant to be approximations. It's a model. All models are false. Some, hopefully, are useful.... with a handful of equations you can describe a dynamic system that kinda, sorta, looks like the vastly more complex real world, and that's useful.

Secondly (and this is more speculative) a corporation need not consciously recognize what it is doing in order to behave in a certain way. To extend DLJ's natural argument, corporate structures could have evolved such that the end result is to behave in a way governed by a few simple equations. To borrow an analogy from the excellent Richard Dawkins: a bat's echo-location system involves the bat squeaking, listening to the echos, and inferring where the obstacles are (all in the presense of thousands of other squeaking bats). That little bat-brain must have some US military grade signal processing built in. Does the bat need to know how it does what it does? No. It just evolved: bats that are able to do this find food, bats that can't do this bump into things and die. Perhaps corporations evolve the same way :)

*(There are two places where economics goes astray, IMHO. Drawing naive and careless welfare implications from that same handful of equations is one, and simple overconfidence... forgetting that these models are only MODELS, is another.)