What will take the con out of econometrics?

Economists Thomas Cooley and Stephen LeRoy are concerned with money demand as an application of econometrics. That applied econometrics is not currently in the most robust of health is hard to deny, and it would be difficult to find as entertaining or as perceptive an analysis of its ills as that found in researcher Edward Learner's various articles. The prescriptions made in those articles are inappropriate, in part because of faulty diagnosis. Extreme bounds analysis is most emphatically not the medicine to cure an ailing patient. This article argues that extreme bounds are generated by the imposition of highly arbitrary restrictions between the parameters of a model. The belief that many of the difficulties applied econometrics currently faces originate in the very poor attempts currently made to accurately describe the process whereby a model was selected, and to ascertain its adequacy. The article proposes a three-stage approach to modeling, involving the selection and subsequent simplification of a general model and a rigorous evaluation of any preferred model. In Cooley and LeRoy's specification the demand for real money is held to be a function of two interest rate variables, the savings and loan passbook rate and the ninety-day Treasury bill rate, real Gross national product, the current inflation rate, the real value of credit card transactions, and real wealt