A0501
Title: Approximate maximum likelihood estimation of a threshold diffusion process with separate drift and diffusion regimes
Authors: Edward MH Lin - Tunghai University (Taiwan)
Henghsiu Tsai - Academia Sinica (Taiwan) [presenting]
Abstract: A novel threshold diffusion model in which the drift and diffusion components undergo regime shifts at distinct threshold values is proposed. Specifically, the drift term changes regimes at one threshold, while the diffusion term transitions at a potentially different threshold level. This doubly thresholded specification offers greater flexibility in capturing asymmetric dynamics in both the conditional mean and volatility. To estimate the model, an approximate maximum likelihood estimation (AMLE) procedure is developed that remains computationally tractable despite the model's structural complexity. Monte Carlo simulations confirm the consistency and efficiency of the proposed AMLE method and demonstrate that standard information criteria (AIC, BIC, HQIC) can reliably distinguish between models with shared versus distinct thresholds. Empirical applications using U.S. Treasury interest rates and the 10-Year/3-Month yield spread further support the proposed framework, revealing structural asymmetries that are better captured by allowing separate thresholds in the drift and diffusion components.
