Forecasting US Output Growth With Mixed Frequency Data

This paper shows forecast results of the US output growth using the leading indicators. The output growth is sampled in quarterly data and the leading indicators are available in monthly frequency. Here I use a mixed frequency regression models to forecast the US output growth at forecast horizons up to one year. This approach is called MIDAS. Two kinds of data vintages are used in this research, the real-time and the end-of-sample vintages. My findings are that the fore-cast accuracy of MIDAS regression is significantly better than the autoregressive model and other competitors. In most cases, using the real-time vintage had better forecast performance.