Palgrave Handbook of Econometrics: Applied Econometrics

David F. Hendry 21

–1.0 –0.5 0.0 0.5 1.0 1.5

1

2

3

Change in log bankruptcies density

1700–1791

1792–1891

1892–1991

–3 –2 –1 0 1 2

1

2

3

4

Change in log patents density

1700–1791

1792–1891

1892–1991

–0.2 –0.1 0.0 0.1 0.2

2.5

5.0

7.5

10.0

Change in log industrial output per capita density

1700–1791

1792–1891

1892–1991

–1.0 –0.5 0.0 0.5

1

2

3

4

5

Change in log real equity prices density

1700–1791

1792–1891

1892–1991

Figure 1.5 Three centuries of data distributions of changes

empirical research should usually involve many variables, although final selections
may prove to be parsimonious (an implication is considered in section 1.5): we now
consider that route.

1.4.2 Incomplete specifications

“What am I to do?” exclaimed Alice, looking about in great perplexity as

first one round head, and then the other, rolled down from her shoulder,

and lay like a heavy lump in her lap. (Lewis Carroll, 1899)

Economies are so high dimensional, interdependent, heterogeneous, and evolv-
ing that a comprehensive specification of all events is impossible: the number of
economy-wide relevant variables is uncountable in a human lifetime. Reducing
that high dimensionality by aggregation over any or all of time, space, com-
modities, agents, initial endowments, etc., is essential, but precludes any claim
to “truth.” So if one cannot get at the “truth,” what is on offer in economics?
Three alternatives are: imposing theory-based models; constructing partial mod-
els, which aim to estimate some parameters associated with a theory, usually by
generalized method of moments (GMM); or seeking the local DGP guided by eco-
nomic theory. All three could operate, but depend on different assumptions. We
first outline how empirical models must arise, then evaluate the three approaches
in general against that basis.