Estimation in Financial Models - RiskLab
Estimation in Financial Models - RiskLab
Estimation in Financial Models - RiskLab
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Introduction<br />
Over the last few years various new derivative <strong>in</strong>struments have emerged <strong>in</strong><br />
nancial markets lead<strong>in</strong>g to a demand for versatile estimation methods for<br />
relevant model parameters. Typical examples <strong>in</strong>clude volatility, covariances<br />
and correlations. In this paper we give a survey on statistical estimation<br />
methods for both discrete as well as cont<strong>in</strong>uous time stochastic models.<br />
The text is organized as follows: <strong>in</strong> Chapter 1 we rst motivate a model <strong>in</strong><br />
which volatility ofa price process is assumed to follow astochastic process.<br />
Out of the variety ofcont<strong>in</strong>uous time stochastic volatility models <strong>in</strong>troduced<br />
<strong>in</strong> the literature wechoose two empirically relevant ones, that is an arithmetic<br />
Ornste<strong>in</strong>-Uhlenbeck process and a square root diusion model. Those two<br />
models serve as reference models <strong>in</strong> some of the later chapters. As for discrete<br />
time stochastic volatility models, we concentrate on log-AR(p) processes,<br />
ARCH(q) and GARCH(p; q) processes all of which will be discussed <strong>in</strong> detail<br />
<strong>in</strong> Section 3.2.<br />
Approximations of diusion models by discrete time models and vice versa<br />
are described <strong>in</strong> Chapter 2. The convergence result on which these approximations<br />
are based is stated. As applications the diusion limits of GARCH(1,1)-<br />
type processes and of AR(1) E-ARCH processes are derived. In addition, we<br />
present strategies for approximat<strong>in</strong>g diusions and briey compare dierent<br />
discretizations of a diusion model.<br />
In Section 3.1 estimation of an unknown parameter <strong>in</strong> a general diusion<br />
process is discussed. For cont<strong>in</strong>uously observed processes we develop the classical<br />
theory of maximum likelihood estimation <strong>in</strong>clud<strong>in</strong>g properties like consistency<br />
and asymptotic normality. In the case of discrete observations the<br />
maximum likelihood estimator reta<strong>in</strong>s all the 'good' properties if the transition<br />
densities of the process are known. However, <strong>in</strong> most cases relevant for<br />
nance, we do not have explicit expressions for the underly<strong>in</strong>g transition densities<br />
and the use of approximate likelihood functions leads to <strong>in</strong>consistent<br />
estimators when the time between observations is bounded away from zero.<br />
We describe alternative estimation methods, as there are mart<strong>in</strong>gale estimat-<br />
3