The fact that the postulates of both human and physical capital have many
observable implications outside the contexts of aggregate models is important
in specific, quantitative ways, in addition to simply giving aggregative theorists
a sense of having 'microeconomic foundations'. For example, in my application
of a human capital model to U.S. aggregative figures, I matched the U.S.
observations to the predictions of a competitive model (as opposed to an
efficient one) in spite of the fact that education, in the U.S., involves vast
government intervention and is obviously not a competitive industry in any
descriptive sense. Why not instead identify the observed paths with the
model's efficient trajectories? The aggregative data have no ability to discriminate
between these two hypotheses, so this choice would have yielded as
good a 'fit' as the one I made. At this point, I appealed to the observation that
most education subsidies are infra-margnal from the individual's point of
view. This observation could stand considerable refinement before it could
really settle this particular issue, but the point is that aggregate models based
on constructs that have implications for data other than aggregates - models
with 'microeconomic foundations' if you like - permit us to bring evidence to
bear on questions of aggregative importance that cannot be resolved with
aggregate theory and observations alone. Without the ability to do this, we can
do little more than extrapolate past trends into the future, and then be caught
by surprise every time one of these trends changes.