A Study of the Effect of Data Transformation and «Linearization» on Time Series Forecasts. A Practical Approach

Very often in actual macroeconomic time series there are causes that disrupt the underlying stochastic process and their treatment is known as «linearization». In addition, variance non-stationarity is in many cases also present in such series and is removed by proper data transformation. The impact of either (data transformation - linearization) on the quality of forecasts has not been adequately studied to date. This work examines their effect on univariate forecasting considering each one separately, as well as in combination, using twenty of the most important time series for the Greek economy. Empirical findings show a significant improvement in forecasts’ confidence intervals, but no substantial improvement in point forecasts. Furthermore, the combined transformation-linearization procedure improves substantially the non-normality problem encountered in many macroeconomic time series