10 Methodology of Empirical Econometric Modeling
progress might be somewhat inefficient when new theories are easily generated as
variants of previous incarnations.
Similar comments apply to Koopmans’ claim that “without resort to theory, in
the sense indicated, conclusions relevant to the guidance of economic policies can-
not be drawn.” Such an argument is not sustainable in general. Originally, aspirin
was a folk remedy for hangovers, derived from brewing willow-tree bark – of which
acetylsalicylic acid, aspirin’s active ingredient, is a natural constituent – without
any theory as to how or why that policy intervention worked (see Weissmann,
1991). A less well known example is the use in folk medicine of fungal-based prod-
ucts, some of which contain natural antibiotics such as penicillin: over 3,000 years
ago, the Chinese had used moldy soybean curd for treating skin infections, again
with no theory on which to base that policy. Theories can be invaluable, and can
enhance the credibility of proposed policies, but they are not essential, especially
when they are incorrect.
1.3.3 The rise of economic theory-based econometrics
“If you’ll tell me what language ‘fiddle-de-dee’ is, I’ll tell you the French
for it!” she exclaimed triumphantly. (Quote from Alice to the Red Queen
in Lewis Carroll, 1899)Another critique of empirical modeling follows from the joint dependence of eco-
nomic events, namely the resulting issues of endogeneity and identification. It
seems widely believed that identification restrictions must be givena prioriby eco-
nomic theory, especially in simultaneous systems, yet that belief also does not have
a substantive basis, as shown in section 1.4.7 on identification.
Together, the cumulative critiques just noted led to an almost monolithic
approach to empirical econometric research: first postulate an individualistic, inter-
temporal optimization theory; next derive a model therefrom; third, find data with
the same names as the theory variables; then select a recipe from the econometrics
cookbook that appropriately blends the model and the data, or if necessary, develop
another estimator; finally report the newly forged empirical law. Thus, we have a
partial answer to the issue posed in section 1.2: the contingent history of econo-
metrics suggested that the only viable route for applied research in economics,
where all current-dated variables are potentially endogenous, was to provide the
quantitative cloth for a completely pre-specified theoretical formulation derived
from general economic principles. But that approach too is problematic and not
without its critics. Economic theory has progressed dramatically over the past
century – imagine being forced to impose 1900’s economic theory today as the
basis for empirical research. If you recoil in horror at that idea, then you have
understood why much past Applied Econometrics research is forgotten: discard
the economic theory that it quantified and you discard the empirical evidence.
Instead of progress, we find fashions, cycles and “schools” in research. The prob-
lem is not that early pioneers ignored economic theory, but that the available
theory was seriously incomplete – as it still is today. Indeed, the Cowles Commis-
sion research was essentially predicated on the belief that the relevant economic