The Methodology of Empirical Econometric Modeling: Applied Econometrics Through the Looking-Glass

This chapter considers the methodology of empirical econometric modeling. The historical background is reviewed from before the Cowles Foundation to the rise of economic theory-based econometrics and the decline of data concerns. A theory for “Applied Econometrics” suggests reinterpreting the role of economic theory given that the intrinsic non-stationarity of economic data vitiates analyses of incomplete specifications based on ceteris paribus. Instead, the many steps from the data-generation process (DGP) through the local DGP (LDGP) and general unrestricted model to a specific representation allow an evaluation of the main extant approaches. The potential pitfalls confronting empirical research include inadequate theory, data inaccuracy, hidden dependencies, invalid conditioning, inappropriate functional form, non-identification, parameter non-constancy, dependent, heteroskedastic errors, wrong expectations formation, misestimation and incorrect model selection. Recent automatic methods help resolve many of these difficulties. Suggestions on the teaching of “Applied Econometrics” are followed by revisiting and updating the “experiment in applied econometrics” and by automatic modeling of a four-dimensional vector autoregression (VAR) with 25 lags for the numbers of bankruptcies and patents, industrial output per capita and real equity prices over 1757–1989.