Essays on Cointegration Analysis

Abstract

Essays in cointegration analysis never was the working title of the work in progress for the last seven years. I started this project with the aim of doing applied research in economics and econometrics. Hence the last chapter of this thesis, Money demand in the Netherlands, was the first chapter written. Yet that very first version, written in Florence in 1998, bears little resemblance to the present version, included in this thesis. The only substantial agreement with the first version are the data used: the short term interest rates were collected from private banks in the Netherlands: as they increasingly offered above money-market interest rates to retail investors, those official interest rates were not relevant for retail investors and even small firms. That interesting problem - the irrelevance of money market rates to money demand - was thus solved rather quickly. Yet the other five chapters all evolved from practical problems, I ran into, during the cointegration analysis of the Dutch data set. I shall thus very briefly describe the cointegrated VAR models and then discuss in turn the problems, these chapters deal with. Each chapter is stand-alone, in that it can be read without having to read any of the others first: this also means that short general overviews of the methodology are presented in each one of them and not repeated here. This introduction just points at the general problems tackled in the papers. For each of the chapters, Matlab programs are available to replicate the results. These are included on a CD, which is part of this thesis. Two programs have been developed into stand-alone packages: 'me2' for maximum likelihood estimation of I(2) models and 'datamine' for the automatic identification and restriction of the cointegration space. Both of them are used in the replication of the results. The replication notes are thus a good starter for using these programs.