Supervision in factor models using a large number of predictors

Project: Research

See relations at Aarhus University

Description

The aim of the project is to investigate the forecasting performance of a particular supervised factor model (SFM). We use a semi-parametric state space representation of the SFM in which the forecast objective, as well as the factors, is included in the state vector.The factors are informed (supervised) of the forecast target through the state equation dynamics. Exploiting the Kalman filter recursions we compute the contribution of the forecast objective to the filtered factors and propose a test to assess its statistical significance. We forecast one target at a time based on the filtered states and estimated parameters of the state space system. We apply the model to time series selected from a set of 132 macroeconomic variables. We compare the forecasting performance of the suggested SFM to benchmark multivariate and univariate models, by means of the conditional predictive ability test of Giacomini and White (2006).
StatusFinished
Effective start/end date01/09/201330/11/2015

ID: 129016961