Aspects of macroeconometric time series modelling

Abstract

This thesis contains six chapters which investigate different areas in applied econometrics. The major focus of the study has been the application of techniques from the applied econometrics literature to a study of the Australian macroeconomy. Chapter Two uses a Vector AutoRegressive (VAR) model and a structural model of the Australian economy to discover those variables responsible for the fluctuations which have buffeted the Australian economy over the last fifteen years. Despite marked differences in the appearance of the two models, the results are similar in predicting how the economy responds to certain shocks. Chapter Three examines the behaviour of the Australian dollar over the period since float in December 1983. The analysis shows that the dollar is over-valued, compared with a level that can maintain a sustainable debt-GDP ratio . The over-valuation has meant that the Australian dollar is discounted on the forward market and high domestic interest rates are necessary to offset the depreciation expected by foreign investors. Chapter Four conducts a Monte Carlo analysis to investigate the performance of alternative estimation methods in equations which include a generated regressor as an explanatory variable. The results show that while FIML tends to dominate with an increasing sample size, in small samples FIML standard errors are downward biased, leaving Correct OLS as the best estimation method. Chapter Five further examines the generated regressor problem using Barro’s (1977) New Classical unemployment model and shows that the results are robust to the estimation method. However, the results from the larger model suggested by Pesaran (1982) are sensitive to the estimation procedure from the larger model suggested by Pesaran (1982) are sensitive to the estimation procedure. Chapter Six evaluates alternative procedures for converting qualitative expectation responses to quantitative expectations for the Australian manufacturing sector and finds that a dynamic nonlinear model which is a generalisation of the model suggested by Pesaran (1987) is superior in picking up both turn in g points in the data and in minimising the forecast error. Chapter Seven further examines the behaviour of the Australian manufacturing sector using multivariate cointegration and the derived quantitative expectations of Chapter Six. The analysis shows that the role of price variables is much more significant than that of output in determining employment movements. Show more