Introduction

We are pleased to have this opportunity to make a contribution to this book in memory of a distinguished Greek econometrician. Michael Magdalinos, from the early years as a PhD student to more recent years when he made important contributions to the literature on hypothesis testing and in particular the role of higher-order approximations to improve inference, impressed the first author by his considerable technical abilities, his tenacity in solving problems, and his warmth as a friend and colleague. Though it was to theoretical econometrics that Michael made his contributions, ultimately the value of many theoretical results lies in applications. We review an issue in applied econometrics in this chapter, namely the efficacy of alternative methods of response analysis in economic policy investigations.

From the 1950s, dynamic simultaneous equations models (DSEM) were a major tool in applied macroeconometric modelling (see for example Klein, Ball, Hazlewood and Vandome, 1961; Klein, 1969), with dynamic multipliers (Goldberger, 1964) being the construct used to measure policy responses. However, following the critiques of economic time-series analysis made by Box and Jenkins (1976) and Granger and Newbold (1974), namely that ignoring non-stationarity in data has serious consequences, time-series models involving differences of data became prevalent. Sims (1980) also criticized DSEMs for employing incredible untested identifying restrictions, and proposed that vector autoregressions (VAR) be used instead. Indeed Sims (1980) was a major influence leading to the VAR becoming the dominant class of model for macroeconometric modelling, and to impulse response analysis being the conventional form of response analysis.