Nowcasting real GDP growth with business tendency surveys data: A cross country analysis

Abstract

We use nowcasting methodology to forecast the dynamics of the real GDP growth in real time based on the business tendency surveys data. Nowcasting is important because key macroeconomic variables on the current state of the economy are available only with a certain lag. This is particularly true for those variables that are collected on a quarterly basis. To conduct out-of-sample forecast evaluation we use business tendency surveys data for 22 European countries. Based on the different dataset and using out-of-sample recursive regression scheme we conclude that nowcasting model outperforms several alternative short-term forecasting statistical models, even when the volatility of the real GDP growth is increasing both in time and across different countries. Based on the Diebold-Mariano test statistics, we conclude that nowcasting strongly outperforms BVAR and BFAVAR models, but comparison with AR, FAAR and FAVAR does not produce sufficient evidence to prefer one over another.