Greek GDP revisions and short-term forecasting

Indicators of economic activity, such as gross domestic product (GDP), are usually published with a significant delay, meaning that central banks and ministries rely on estimates or predictions of the key economic indicators in order to conduct monetary and fiscal policy. The econometric method-ology that is commonly used to provide a timely estimation of the current state of the economy is referred to as nowcasting and is based on the use of economic indicators that are published earlier and at a higher frequency than the target variable. This study focuses on the Greek economy and particularly on Greek GDP and examines the effect of GDP data revisions on the out-of-sample fore-casting outcome of alternative nowcasting models, by utilising real-time GDP data. To that end, we construct a real-time GDP database and we compare the predictive ability of alternative nowcasting models using both last vintage and real-time databases. The empirical results for an out-of-sample period of ten years (2007-2017) show that usually a model with a small set of real variables and the PMI can consistently produce good GDP forecasts as we move closer to the GDP publication date; most importantly, this result is not affected by the GDP revisions and holds true using both the last vintage GDP and the real-time GDP data.