A1572
Title: Nowcasting with signature methods
Authors: Lingyi Yang - Alan Turing Institute (United Kingdom)
Aureo de Paula - University College London (United Kingdom)
Lars Nesheim - University College London (United Kingdom)
Samuel Cohen - University of Oxford (United Kingdom)
Giulia Mantoan - Alan Turing Institute (United Kingdom)
Silvia Lui - Office for National Statistics (United Kingdom)
Emma Small - Office for National Statistics (United Kingdom)
Craig Scott - Office for National Statistics (United Kingdom)
Will Malpass - Office for National Statistics (United Kingdom)
Giulia Mantoan - The Alan Turing Institute (United Kingdom) [presenting]
Abstract: Nowcasting refers to the ``forecast'' of the current (``now'') state of the economy. This is necessary as key economic variables are often published with a significant delay of over a month. The nowcasting literature has arisen to address the need to have fast, reliable estimates of delayed economic indicators. The path signature is a mathematical object which captures geometric properties of sequential data; it naturally handles missing data from mixed frequency and/or irregular sampling - issues often encountered when merging multiple data sources - by embedding the observed data in continuous time. Calculating path signatures and using them as features in models have achieved state-of-the-art results in other fields such as finance, medicine, and cyber security. We look at the nowcasting problem by applying regression on signatures, a simple linear model on these nonlinear objects that we show subsumes the popular Kalman filter. We quantify the performance via a simulation exercise and application to US GDP growth, where in the latter, we compare performance with the dynamic factor model. By embedding discrete information in continuous time, this approach allows greater flexibility for future applications on data with complex sampling patterns.