Volatility modeling in the presence of measurement errors

Jonas Andersson and Anders Ågren

Abstract

ABSTRACT

In this paper, the authors examine the effects of measurement errors on volatility measures. This is done by first expressing the moment properties of general volatility models with measurement errors in terms of the corresponding moments for the underlying unobserved signal process. Then the consequences of measurement errors for some GARCH and stochastic volatility models are evaluated. It is shown that measurement error causes the autocorrelation function, here for the squared process, to start and remain lower than for the underlying unobserved signal process. The size of the effects are highly dependent on the degree of persistence in the volatility. Finally, the consequences of measurement errors on time-varying VaR measures are studied.