B1038
Title: Sufficient directional forecasting using factor models
Authors: Wei Luo - Zhejiang University (China) [presenting]
Lingzhou Xue - Penn State University (United States)
Jiawei Yao - Princeton University (United States)
Abstract: Factor models have been widely used for forecasting in economical research. Instead of the traditional regression of the target response variable on the factors, sliced inverse regression has been applied to further reduce the dimension of the factors, which enhances the accuracy of forecasting while avoids the risk of model mis-specification. We apply directional regression for dimension reduction on the factors, which is more comprehensive than sliced inverse regression. In addition, we allow the dimension of the factors to increase with the sample size. Consequently, compared to the previous approach, ours allows a larger number of factors in finite samples, and is applicable for a more general relationship between the factors and the target response variable. We develop relative asymptotic theory, and conduct simulation studies and real data analysis to illustrate the effectiveness of the proposed method.
Journal