Three essays in applied multivariate econometrics

The objective of this dissertation is firstly to use recently developed econometric techniques to study two relevant economic problems, and secondly to examine the robustness of one of the techniques used to mean shifts. In the first essay the Hendry and Mizon methodology is used to study the relationship between inflation and relative price variability for the UK using an information set which includes other relevant macroeconomic variables. A structural econometric model is estimated for I(0) variables, including ECMs estimated using the Johansen procedure. We conclude that there is a positive relationship between inflation and relative price variability in the long-run. The second essay consists of an empirical study of the Fisher Hypothesis for the UK. Using both univariate and multivariate approaches, we conclude that inflation does not have a systematic influence on stock returns. In the third essay we study the impact of mean shifts on the Engle and Granger and Johansen procedures. A simple bivariate model is used to test the impact of both impulse and step dummies on these procedures. We find that the inclusion of a dummy, particularly a step dummy, at the estimation stage of either procedure, tends to lead to an over-rejection of cointegration. We also find that when using the Engle and Granger procedure, the inclusion of a dummy in the DGP of one variable tends to lead to an under-rejection of cointegration.