14
THE STATISTICAL VERIFICATION OF
Wesley Mitchell discusses whether the view, held by
many statistical economists to-day, that the function of
statistics is to provide a statistical complement for
economics and not to recast economic theory, is sound.
He bases his preference for the most radical suggestion
on a very suggestive consideration of what has happened
in physics.
¢ The mechanical view involves the notions of same-
ness, of certainty, of invariant laws ; the statistical
view involves the motions of variety, of probability,
of approximations. Yet Clerk-Maxwell’s * new kind
of uniformity > was found to yield results in many
physical problems which corresponded closely to results
attained on mechanical lines.
“Such a close correspondence between the results
based on speculation and the results based on statisti-
cal observation is not to be expected in economies, for
three reasons. First, the cases summed up in our
statistics seldom if ever approach in number the mil-
lions of millions of molecules, or atoms, or electrons
of the physicist. Second, the units in economic aggre-
gates are less similar than the molecules or atoms of
a given element. Third, we cannot approach closely
the isolation practices of the laboratory. For these
reasons the elements of variety, of uncertainty, of
imperfect approximation are more prominent in the
statistical work of the social sciences than in the
statistical work of the natural sciences.
¢ And because our statistical results are so marked
by these imperfections they do not approach so closely
to the results of our reasoning on the basis of assumed
premises. Hence the development of statistical method
may be expected to make more radical changes in
economic than it makes in physical theory.’
The analysts reach their mass generalizations about
markets and demand by considering the individual’s
behaviour and multiplying him into results which can
never be precise, whereas the statistician slips the in-
dividual stage and treats of masses and modes direct.
