14 Methodology of Empirical Econometric Modeling
We now consider each of the twelve assumptions in turn, devoting the separate
section 1.5 to the last, namely model selection.
1.4.1 Economic theory
“It seems very pretty,” she said when she had finished it, “but it’s rather
hard to understand.” (Alice after reading the Jabberwocky poem in Lewis
Carroll, 1899)Economic theory has created many major ideas that have in turn changed the
world, from the “invisible hand” in (Smith, 1759, p. 350), understanding the gains
from trade and the problems with mercantilism, through the effects of tariffs and
taxes, to modern insights into issues such as welfare economics, option pricing,
auctions, contracts, principal–agent and game theories, trust and moral hazard,
asymmetric information, institutions, and all their attendant impacts on market
functioning and industrial, and even political, organization. In doing so, eco-
nomics has evolved dramatically over time, and will undoubtedly continue doing
so, hence at no instant can it be claimed to be correct, complete and immutable.
For example, most theories take preferences as a given – sometimes even as “deep
parameters” – but there are many endogenous determinants, from learning, adap-
tation, and advertising among others (see, e.g., von Weizsacker, 2005), with
psychological, behavioral, and neuro-economics bidding fair to play key roles in
the future (see,inter alia, Fehr and Falk, 2002; Fehr, Fischbacher, Kosfeld, 2005;
Camerer, 2007).
Theories need to be distinguished in terms of their “levels’, where low-level
theories are well established and widely accepted (e.g., the optical theory behind
the design of telescopes and the interpretation of their evidence), whereas high-
level theories usually assume the validity of many lower levels, but are subject to
doubt (as in theories of the accelerating expansion of the universe as due to “dark
energy”). Facts are items of empirical information which depend only on low-level
theories and measurements, and can be reliably replicated. Since all empirical evi-
dence is theory laden to some degree, albeit often just from very low-level theories,
“measurement without theory” is trivially impossible, and must relate to the lack of
use of high level theories – the appropriate blend of theory and empirical evidence
affects research efficiency, not necessarily the validity of any resulting findings
(see, e.g., Gilbert and Qin, 2007). Many low-level statements are correct, complete
and immutable, such as 1+ 1 =2, and although essential to arithmetic, cannot
“explain” economic behavior. Conversely, testing theories just by their predictions
is problematic, as false assumptions can entail correct conclusions: assume 1=2,
then 2=1, so adding both sides, 3=3, which is valid and presumably thereby
establishes that 1=2 (also see Ericsson and Hendry, 1999).
General theories that do “explain” theGestaltof empirical evidence are a boon,
but are not essential. Similarly, although experimentation can be helpful, it is far
from the only source of evidence: observational science is not an oxymoron. The
progressivity of science – its cumulation of findings that cohere, are consolidated in
theoretical explanations, and suggest where next to investigate – is its most salient