A1183
Title: Inference for the panel ARMA--GARCH model when both N and T are large
Authors: Bing Su - The University of Hong Kong (China) [presenting]
Ke Zhu - University of Hong Kong (Hong Kong)
Abstract: A panel ARMA-GARCH model is proposed to capture high-order serial correlation and conditional heteroskedasticity of large panel data with N individuals over T time periods. For this model, a two-step estimation procedure is provided to estimate the ARMA parameters and GARCH parameters sequentially. Under some regular conditions, the proposed estimators are shown to be asymptotically normal with convergence rate $(NT)^{-1/2}$, and they have asymptotic biases when both N and T diverge jointly to infinity at the same rate. Particularly, the asymptotic biases result from the fixed effect, estimation effect, and unobservable initial values. Furthermore, either analytical asymptotics or the jackknife method is proposed to construct bias-corrected versions of the estimators, and the validity of both methods is justified. Additionally, the panel ARMA-GARCH model is generalized to incorporate a latent group structure, accommodating parameter heterogeneity across groups. Finally, the usefulness of the panel ARMA-GARCH model and its generalization is demonstrated through simulations and empirical applications.
