OPTION PRICING UNDER THRESHOLD AUTOREGRESSIVE MODELS BY THRESHOLD ESSCHER TRANSFORM

This paper develops a valuation model for options under the class
of self-exciting threshold autoregressive (SETAR) models and their variants for
the price dynamics of the underlying asset using the self-exciting threshold autoregressive Esscher transform (SETARET). In particular, we focus on the first
generation SETAR models first proposed by Tong (1977, 1978) and later developed in Tong (1980, 1983) and Tong and Lim (1980), and the second generation
models, including the SETAR-GARCH model proposed in Tong (1990) and the
double-threshold autoregressive heteroskedastic time series model (DTARCH)
proposed by Li and Li (1996). The class of SETAR-GARCH models has the
advantage of modelling the non-linearity of the conditional first moment and
the varying conditional second moment of the financial time series. We adopt
the SETARET to identify an equivalent martingale measure for option valuation in the incomplete market described by the discrete-time SETAR models.
We are able to justify our choice of probability measure by the SETARET
by considering the self-exciting threshold dynamic utility maximization. Simulation studies will be conducted to investigate the impacts of the threshold
effect in the conditional mean described by the first generation model and that
in the conditional variance described by the second generation model on the
qualitative behaviors of the option prices as the strike price varies.