READ-GER: Introducing German Real-Time Regional Accounts Data for Revision Analysis and Nowcasting

Accurate real-time macroeconomic data are essential for policy-making and economic nowcasting. In this paper, I introduce a real-time database for German regional economic accounts (READ-GER). The database contains real-time information for nine macroeconomic aggregates and the 16 German states. I conduct both a revision analysis and a nowcasting experiment. Whereas the first estimates show no systematic revision errors by pooling the states together, this procedure suppresses the revision characteristics of single states. For half of the 16 German states I find that the first estimates are no optimal predictions, thus, leaving room for improvements in the future. The real-time nowcasts for real gross domestic product growth based on a mixed-frequency Vector Autoregression are very accurate, beat several benchmark models and are as precise or better as the first official estimates. More regional data would help to further increase the model's nowcast performance and thus its properties for the first estimates from regional accounts.