skip to content

Faculty of Economics


Cristea, R. G.

Can Alternative Data Improve the Accuracy of Dynamic Factor Model Nowcasts?


Abstract: We take the standard dynamic factor model for euro area real GDP growth nowcasting and test how adding several extensions improves forecasting precision. We expand the model's information set with high frequency alternative data and amend how some of the traditional variables are considered. Subsequently, we enrich the factors structure with blocks for soft data, labour and financial markets, real-time data and the supply side of the economy. As a result, our enriched nowcast has accurately detected the downturn in Q1-2020 and has correctly indicated further and steeper contraction in Q2 due to the COVID-19 shock. Results from several genuine and pseudo real-time out-of-sample forecast evaluation exercises show nowcasting precision gains, as measured by the root mean squared error or Kuiper's score. We also show these gains stem from the novel data sources and factors structure. While our model's outperformance is modest in normal times, it is meaningful in times of severe stress.

Keywords: nowcasting, dynamic factor model, Kalman filter, real-time high frequency alternative data, Google econometrics, COVID-19, euro area macroeconomics

JEL Codes: C38 C53 C55 E27 E66 Y40

Author links: Radu Cristea  


Open Access Link: