How to forecast wage and price inflation with more accuracy: the Australian experience 1949/50 to 1988/89

Econometric estimates and forecasts based on a large number of current wage and price inflation models fitted to Australian pre- and post- 1983 Accord data have not been satisfactory (see for example Chapman, 1990). The unsatisfactory results may be due to poor proxy variables, misspecified models, incorrectly measured data or inappropriate modelling methods. This paper focuses on the general methodological issue and provides, for illustration, improved estimates and forecasts of a Dicks-Mireaux and Fels-Tran Van Hoa model of wage and price inflation in Australia for the period 1949/50 to 1988/89. Based on our new improved methods, a net gain in accuracy of up to 97 percent in terms of stochastic simulation MSE and up to 87 percent in terms of ex post forecasting MSE (Pindyck and Rubinfeld, 1991) is obtained.