12-FinEcon-Model Sel Page 315 Wednesday, February 4, 2004 12:59 PM

I

CHAPTER

12

Financial Econometrics:

Model Selection, Estimation,

and Testing

n economics and finance theory models are rarely determined by

strong theoretical considerations. Often, one or more families of mod-

els compete as plausible explanations of empirical data. Therefore, a

specific family of models has to be selected and, within a given family,

parameters have to be estimated. In this chapter we discuss criteria for

model selection and parameter estimation.

MODEL SELECTION

Science works by making hypotheses and testing them. In the physical

sciences, in particular, hypotheses are mathematical models typically

tested with a very high level of precision under a variety of experimental

settings. In the usual process of scientific inquiry, models can be under-

stood as the product of human creativity. How the general concepts of

science are formed and modified to account for new empirical evidence

has been the subject of intense study.^1

With the advent of fast computers, an automatic approach to sci-

ence—and to the creative process in general—has been made possible.

The Nobel laureate Herbert Simon was a strong advocate of the idea

that the creative discovery process can be automated as an algorithmic

(that is, step-by-step) search in a space of different possibilities.

(^1) See for instance Thomas Kuhn, The Structure of Scientific Revolutions: Third Edi-
tion (Chicago: University of Chicago Press, 1996).
315