The Nowcasting Lab: Live Out-of-Sample Forecasting and Model Testing

Philipp Kronenberg,Heiner Mikosch,Stefan Neuwirth, Matthias Bannert, Severin Thöni

SSRN Electronic Journal(2023)

引用 1|浏览0
暂无评分
摘要
The Nowcasting Lab is an automated code-database-website environment for GDP forecasting. It generates nowcasts and one-quarter ahead forecasts for quarterly GDP growth of the United States, the euro area, and currently 14 other economies using several forecasting models and a large amount of data. The predictons are updated daily and released on a website together with detailed additional information. Forecasting practitioners can use the website to support their own work. Researchers can use the lab to monitor and test the performance of forecasting models in a live out-of-sample environment. All predictions and input data are stored in a daily vintage database which can be used for real-time forecasting studies. As an application, we analyze the live out-of-sample now- and forecast performance of four mixed-frequency forecasting models during the COVID-19 crisis. The models failed to predict the large fluctuations in GDP growth in the year 2020. For the year 2021, there are large differences in prediction accuracy depending on the considered economy and model used.
更多
查看译文
关键词
model testing,nowcasting lab,out-of-sample
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要