VG NGARCH versus GARJI model for asset price dynamics

This study proposes and calibrates the VG NGARCH model, which provides a more informative and parsimonious model by formulating the dynamics of log-returns as a variancegamma (VG) process by Madan et al. (1998). An autoregressive structure is imposed on the shape parameter of the VG process, which describes the news arrival rates that affect the price movements. The performance of the proposed VG NGARCH model is compared with the GARCH-jump with autoregressive conditional jump intensity (GARJI) model by Chan and Maheu (2002), in which two conditional independent autoregressive processes are used to describe stock price movements caused by normal and extreme news events, respectively. The comparison is made based on daily stock prices of five financial companies in the S&P 500, namely, Bank of America, Wells Fargo, J. P. Morgan, CitiGroup, and AIG, from January 3, 2006 to December 31, 2009. The goodness of fit of the VG NGARCH model and its ability to predict the ex ante probabilities of large price movements are demonstrated and compared with the benchmark GARJI model.