Complete "realism" is clearly unattainable, and the question whether a
theory is realistic "enough" can be settled only by seeing whether it yields
predictions that are good enough for the purpose in hand or that are better
than predictions from alternative theories. Yet the belief that a theory can be
tested by the realism of its assumptions independently of the accuracy of its
predictions is widespread and the source of much of the perennial criticism of
economic theory as unrealistic.-Milton
Friedman, The Methodology of Positive Economics, 1953.
My friends with engineering and
natural science background often fall into a state of cognitive dissonance when
I try to explain to them how models in economics actually work and why very
profound and useful results can be obtained from models with conditions that
seem to be counter-intuitive, esoteric and "not realistic". Why subjective
future discount factor is set at magical number of 0.96? Why would venture capital and private equity
managers typically get 20% share of enterprise profit or more but never less
nor matter how bad the state of economy is? Why an advanced model with limited
number of highly simplified agents and production sectors but with private information
and capital misallocations actually works well while complex multi-sector model
with detailed inputs for each factor of production done in standard CGE
framework according to the best mathematical and computational recipes fails
miserably to describe even the simplest price and output dynamics anywhere but
the most trivial cases?
At some level general equilibrium theory is vacuous. For example, for
pure exchange economies it is known that one can generate any aggregate excess
demand function by a suitable specification of endowments and preferences.
Related, unobserved shocks to preferences with arbitrary probability
distributions can generate arbitrary patterns of cross household consumptions.
So, if some version of an enlarged model
always fits, to what extent does general equilibrium theory have content?
...
Again, general equilibrium models allow one to think logically about
the implications of some premise, to trace out the implications of the premise
not only for the phenomenon of interest but for other phenomena as well. That
is, one can keep track of all possible interactions across agents in the
economy, showing some phenomena to be
logically inconsistent with other phenomena.
-Robert M. Townsend,"Models as Economies" in The Economic Journal, 98
(1987).
Try to imagine the physical world
in which gravitational "constant" is no longer constant but changes from place
to place and unexpectedly in time. It would take quite a leap of imagination to
discover a "true" law of motion in such world when so much visible complexity
is lying on the surface and so many alternative versions would fit "reality"
equally well. And that's precisely the word where economics models apply. It is
really amazing that there exist forms of economic activity including their
quantitative characteristics (such as ex-ante division of profits in
entrepreneurial contracts) that can be observed and traced through hundreds of
years in human history. To find such economic patterns and more importantly the
dynamics of change in those patterns both spatially and across time you need to
go into a fairy land of economic modeling with strict logical, quantitative,
computational discipline enforced and come back again to a subset of actual
economies and data in which relevant phenomena manifest themselves most
clearly. This journey of course is fraught with many obstacles and perils but
that's what any good journeys is ultimately about.
Post submitted by
Victor Zhorin, University of Chicago Computation Institute
[]
[]