A0207
Title: Instrumental factor models for high-dimensional functional data
Authors: Young-Kwang Kim - Toulouse School of Economics (Korea, South) [presenting]
Jihyun Kim - Sung Kyun Kwan University (Korea, South)
Abstract: The instrumental factor model is introduced that extends conventional factor models in two directions. First, the factor model is developed for high-dimensional data, from scalar-valued data to functional data that has gained fast-growing popularity. Second, it is well-known that the conventional estimation approach based on the principal component analysis (PCA) requires both cross-sectional dimension and the time horizon of data to be large. Under the proposed approach that utilizes additional characteristic variables as instruments, the estimators achieve consistency as long as the cross-sectional dimension is large enough. The eigenuses value ratio method is then proved consistently aided to estimate the unknown number consistently. The numerical experiments confirm that the estimation approach outperforms the conventional PCA-based method, especially for short panel da a. In concussion, with an empirical application to analyze the long-run relationship between global warming and world GDP. The results support the growing consensus that human activities are the dominant cause of global warming.
