Parametric Estimation of Jump-Diffusions（519-530）.pdfVIP

Financial Econometrics 24 Parametric Estimation of Jump-Diffusions* Key words: maximum likelihood estimation, jump-diffusions, Monte Carlo simulation, term structure of interest rates We consider the problem of estimating the coefficients of diffusion processes with jumps. These processes have been widely employed for describing sudden changes both in stock price and in interest rate dynamics and reproducing high-order mo- ments of sample data. In most cases, the lack of analytical expressions for transition densities makes the use of maximum likelihood methods a very challenging task. This case describes and tests the simulated maximum likelihood (SML) method introduced by Pedersen (1995) and Brandt and Santa Clara (2002). This technique allows for the estimation of a wide variety of diffusion processes, including those which lack closed-form expressions for the transition density. We illustrate the SML method by developing algorithms for the estimation of both continuous and mixed- jump-diffusion processes. The latter process has been employed in interest rate mod- elling by Piazzesi (2001). The main idea underlying SML is to numerically evaluate the transition proba- bilities of the process corresponding to all pairs of values taken by the state variable at consecutive times. If a discretization of the time–space axes is properly refined, the resulting transition density approaches a Gaussian distribution. The likelihood estimator becomes a reliable approximation of the exact maximum likelihood esti- mator, namely the one stemming from the exact, yet unknown, transition density of the process. Sections 24.1 and 24.2 introduce the estimation problem and illustrate the gen- eral methodology, respectively. Section 24.3 details the algorithm, while Sect. 24.4 provides two applications. First, we consider the Cox, Ingersoll and Ross (1985) model (CIR) for short-term interest rate dynamics. Since the transition density