6 Methodology of Empirical Econometric Modeling
“theory of Applied Econometrics” which highlights some of the problems empirical
modeling confronts in a non-stationary environment, where non-stationary is used
throughout in the “wide sense” to denote any changes in the distributions of the
random variables modeled by economists. Section 1.5 discusses a recent tool for
automatic modeling, Autometrics, based on the last decade of research into model
selection (see Doornik, 2007a; Hendry and Krolzig, 2005; Hendry, with Doornik
and Nielsen, 2007). Section 1.6 comments on teaching Applied Econometrics, and
section 1.7 revisits the experiment in applied econometrics conducted by Magnus
and Morgan (1999). Section 1.8 then looks at automatic modeling of a four-variable
dynamic system related to industrial output since 1700, with 25 lags in its initial
formulation and many outliers over more than 250 years. Section 1.9 concludes.
Throughout, I draw heavily on a number of my previous papers. Despite there
being almost 200 citations to other scholars, I am conscious that documentation
is bound to be incomplete, and apologize for omitting many contributions.
1.2 What is “Applied Econometrics”?
“When I use a word,” Humpty Dumpty said in rather a scornful tone,
“it means just what I choose it to mean – neither more nor less.” (Lewis
Carroll, 1899)At the superficial level, “Applied Econometrics” is “any application of economet-
rics,” as distinct from theoretical econometrics. If it were not for the imperialist
tendencies of econometricians, that would suffice, but econometrics has been
applied in space science, climatology, political science, sociology, epidemiology,
marketing,inter alia, not to mention the claim inHow the Laws of Physics Lie(see
Cartwright, 1983) that econometrics is the key methodology for all of science...
Sorry to disappoint the eager reader, but I will not be covering even a wide range
of the economic applications, never mind that plethora of outside studies.
Some applied econometricians would include any applications involving anal-
yses of “real economic data” by econometric methods, making “Applied Econo-
metrics” synonymous with empirical econometrics. However, such a view leads
to demarcation difficulties from applied economics on the one hand and applied
statistics on the other. Defining “econometrics,” as in Frisch (1933), to comprise
only studies involving the unification of economic theory, economic statistics
(data), and mathematics (statistical methods) helps in demarcation, but limits its
scope and inadvertently excludes (say) developing econometric theory itself, or
just improving data measurement and collection.
Outsiders might have thought that “Applied Econometrics” was just the appli-
cation of econometrics to data, but that is definitely not so; virtually no journal
editor would publish such a piece. Rather, the notion of mutual penetration domi-
nates – but as a one-way street. Economic theory comes first, almost mandatorially.
Perhaps this just arises from a false view of science, namely that theory precedes
evidence, even though, apart from a few famous occasions, science rarely proceeds
by imposing a preconceived theory on evidence, and evidence regularly shapes and