Stationarity and non-stationarity : issues and implication in econometric forecasting modelling / Mohd Alias Lazim

Over the last decade the concept of stationarity has occupied the mind of many econometricians. This new understanding of time series data has led many to rethink the procedure of forecasting modelling on the evidence that models that seem to fit well are not necessarily well
specified. Hence, in many new applied works, the forecasting
performances of many econometric models have been shown to improve by taking into account the stationarity and non-stationarity characteristics of the variables involved. Amongst the many new developments in the econometric forecasting modelling are models being developed using differenced series. Hence, this paper seeks to explain the concept of stationarity and non-stationarity in economic
time series data. The importance of giving due considerations to these properties when estimating econometric forecasting models is stressed and it is hoped that this discussion may act as a guide to economic
forecasters. Also discussed is the method commonly used to test for non-stationarity.