Some properties of absolute returns as a proxy for volatility

Abstract

We use the stochastic volatility model as a basis for investigating the statistical properties of absolute returns as a measure of latent volatility in financial markets. Our results are compared with existing results for squared returns.

Notes

¹Other measures include the daily price range – e.g. Ghysels et al. (2006).

²The appeal of the SV model over the GARCH model in this context is that it allows the time-dependence of volatility to be stochastic rather than deterministic.

³The independence assumption can be relaxed without affecting the main results below if appropriate conditioning arguments are used.

⁴In addition, Guégan and Diebolt (1994) consider the so-called β-ARCH models that include the special case of predicting future conditional variance through past absolute returns.