Autoregressive Conditional Parameter Model and Applications
32 Pages Posted: 17 Mar 2016 Last revised: 17 Aug 2017
Date Written: July 17, 2016
Abstract
This paper proposes a new time-varying parameter regression called the autoregressive conditional parameter (ACP) model. The model allows for time-varying conditional parameters such as the regression coefficients (beta coefficients), mean/intercept, variance, duration, skewness and kurtosis. In particular, the model provides a new method for time-varying betas. It’s shown that the model encompasses many wide-used time series models (e.g., the rolling linear regression, autoregressive and moving average, generalized autoregressive conditional heteroskedasticity, autoregressive conditional duration, autoregressive conditional skewness and kurtosis, and Poisson autoregression). Identification, estimation and test of ACP model are also examined. The empirical results show that the ACP model generally outperforms two popular time-varying parameter models (i.e., the state space model and dynamic conditional beta model).
Keywords: time-varying conditional parameter, autoregressive conditional parameter model, autoregressive conditional beta, commodity market
JEL Classification: C22, C50
Suggested Citation: Suggested Citation