Title: Extracting factors from large datasets
Authors: Alessia Paccagnini - University College Dublin (Ireland) [presenting]
Abstract: Factor models have become very popular in both shock identification and forecasting analysis. One crucial aspect is to identify the number of factors which summarizes information of the disposable dataset. Several studies propose different methods to estimate the number of factors. In addition, the recent literature about big data has stressed the importance of the increase of the sample of datasets which could create new challenges in this topic. An analysis of different methodologies to estimate the number of factors is proposed, focusing on the increase of the dataset in both time and cross-sectional dimensions. The time variation is taken into consideration. Empirical exercises on the US economy are presented to show how the number of factors matters in shock identification and forecasting analysis.