Séminaire d’économétrie

14h – 16h

jeudi 17 septembre 2026

Salle 3001

Eric GHYSELS (UNC Chapel Hill) “Seasonality in High Dimensional Data”

Résumé :

We propose a novel approach to deal with seasonality in high-dimensional high-frequency data. Our modeling innovation involves two large panels. The first consists of low frequency (say quarterly or monthly) seasonally adjusted series. The second panel contains seasonally unadjusted series where the sheer number of series and their typically recalcitrant and messy seasonal patterns make it difficult to seasonally adjust them. Both panels are high-dimensional, and by construction only share non-seasonal factors.
The constituents of both panel can be very different, but all the series relate to some common sources of fluctuations. The key insight of our novel modeling approach is that we estimate the factor(s) that are common between the two panels. In the empirical application we study commonly used monthly seasonally adjusted macroeconomic series and daily/weekly unadjusted series related to economic activity.

Travail commun : Elena ANDREOU, Patrick GAGLIARDINI, Mirco RUBIN