This paper employs an extension of the mixed-frequency collapsed dynamic factor model, introduced by Bräuning and Koopman (2014), for nowcasting Israeli GDP growth. In the proposed method, the dimension reduction is performed through a partial least squares regression that uses a proxy for the unobserved monthly GDP growth as the dependent variable. Our results show that this method allows us to improve on the accuracy of quarterly forecasts. In addition, we study the role of business surveys in real-time nowcasting. We find that including business surveys improves the accuracy of nowcasting, but only when they are used for endpoint imputation of the traditional macroeconomic indicators. In contrast, expanding the monthly panel with surveys related to some of the already included variables leads to inferior forecasting performance. Finally, the proposed model allows us to construct a monthly index of real economic activity that is consistent with the nowcast. Compared to the Composite State of the Economy Index currently published by the Bank of Israel, it utilizes a much broader data set and thus is likely to provide a more timely and precise picture of the course of economic activity.


Keywords: Nowcasting; Dynamic factor model; Partial least squares.


For a PDF of the full article ​